~dkuhlman/python-training-materials/Materials

Loggers are plain Python objects. The <a class="reference internal" href="../library/logging.html#logging.Logger.addHandler" title="logging.Logger.addHandler"><code class="xref py py-meth docutils literal">addHandler()</code></a> method has no

246

minimum or maximum quota for the number of handlers you may add. Sometimes it

247

will be beneficial for an application to log all messages of all severities to a

248

text file while simultaneously logging errors or above to the console. To set

249

this up, simply configure the appropriate handlers. The logging calls in the

250

application code will remain unchanged. Here is a slight modification to the

251

previous simple module-based configuration example:

252

<div class="highlight-python3"><div class="highlight"><pre>import logging

253

254

255

256

# create file handler which logs even debug messages

257

258

259

# create console handler with a higher log level

260

ch = logging.StreamHandler()

261

262

# create formatter and add it to the handlers

263

264

ch.setFormatter(formatter)

265

fh.setFormatter(formatter)

266

# add the handlers to logger

267

logger.addHandler(ch)

268

logger.addHandler(fh)

269

270

# 'application' code

271

logger.debug('debug message')

272

logger.info('info message')

273

logger.warn('warn message')

274

logger.error('error message')

275

logger.critical('critical message')

276

</pre></div>

277

</div>

278

Notice that the ‘application’ code does not care about multiple handlers. All

279

that changed was the addition and configuration of a new handler named fh.

280

The ability to create new handlers with higher- or lower-severity filters can be

281

very helpful when writing and testing an application. Instead of using many

282

<code class="docutils literal">print</code> statements for debugging, use <code class="docutils literal">logger.debug</code>: Unlike the print

283

statements, which you will have to delete or comment out later, the logger.debug

284

statements can remain intact in the source code and remain dormant until you

285

need them again. At that time, the only change that needs to happen is to

286

modify the severity level of the logger and/or handler to debug.

287

</div>

288

289

<h2>Logging to multiple destinations<a class="headerlink" href="#logging-to-multiple-destinations" title="Permalink to this headline">¶</a></h2>

290

Let’s say you want to log to console and file with different message formats and

291

in differing circumstances. Say you want to log messages with levels of DEBUG

292

and higher to file, and those messages at level INFO and higher to the console.

293

Let’s also assume that the file should contain timestamps, but the console

294

messages should not. Here’s how you can achieve this:

295

<div class="highlight-python3"><div class="highlight"><pre>import logging

296

297

# set up logging to file - see previous section for more details

298

299

format='%(asctime)s %(name)-12s %(levelname)-8s %(message)s',

300

datefmt='%m-%d %H:%M',

301

filename='/temp/myapp.log',

302

filemode='w')

303

# define a Handler which writes INFO messages or higher to the sys.stderr

304

console = logging.StreamHandler()

305

console.setLevel(logging.INFO)

306

# set a format which is simpler for console use

307

formatter = logging.Formatter('%(name)-12s: %(levelname)-8s %(message)s')

308

# tell the handler to use this format

309

console.setFormatter(formatter)

310

# add the handler to the root logger

311

logging.getLogger('').addHandler(console)

312

313

# Now, we can log to the root logger, or any other logger. First the root...

314

logging.info('Jackdaws love my big sphinx of quartz.')

315

316

# Now, define a couple of other loggers which might represent areas in your

317

# application:

318

319

logger1 = logging.getLogger('myapp.area1')

320

logger2 = logging.getLogger('myapp.area2')

321

322

logger1.debug('Quick zephyrs blow, vexing daft Jim.')

323

logger1.info('How quickly daft jumping zebras vex.')

324

logger2.warning('Jail zesty vixen who grabbed pay from quack.')

325

logger2.error('The five boxing wizards jump quickly.')

326

</pre></div>

327

</div>

328

When you run this, on the console you will see

329

<div class="highlight-python3"><div class="highlight"><pre>root : INFO Jackdaws love my big sphinx of quartz.

330

myapp.area1 : INFO How quickly daft jumping zebras vex.

331

myapp.area2 : WARNING Jail zesty vixen who grabbed pay from quack.

332

myapp.area2 : ERROR The five boxing wizards jump quickly.

333

</pre></div>

334

</div>

335

and in the file you will see something like

336

<div class="highlight-python3"><div class="highlight"><pre>10-22 22:19 root INFO Jackdaws love my big sphinx of quartz.

337

10-22 22:19 myapp.area1 DEBUG Quick zephyrs blow, vexing daft Jim.

338

10-22 22:19 myapp.area1 INFO How quickly daft jumping zebras vex.

339

10-22 22:19 myapp.area2 WARNING Jail zesty vixen who grabbed pay from quack.

340

10-22 22:19 myapp.area2 ERROR The five boxing wizards jump quickly.

341

</pre></div>

342

</div>

343

As you can see, the DEBUG message only shows up in the file. The other messages

344

are sent to both destinations.

345

This example uses console and file handlers, but you can use any number and

346

combination of handlers you choose.

347

</div>

348

349

<h2>Configuration server example<a class="headerlink" href="#configuration-server-example" title="Permalink to this headline">¶</a></h2>

350

Here is an example of a module using the logging configuration server:

351

<div class="highlight-python3"><div class="highlight"><pre>import logging

352

import logging.config

353

import time

354

import os

355

356

# read initial config file

357

logging.config.fileConfig('logging.conf')

358

359

# create and start listener on port 9999

360

t = logging.config.listen(9999)

361

t.start()

362

363

364

365

try:

366

# loop through logging calls to see the difference

367

# new configurations make, until Ctrl+C is pressed

368

while True:

369

logger.debug('debug message')

370

logger.info('info message')

371

logger.warn('warn message')

372

logger.error('error message')

373

logger.critical('critical message')

374

time.sleep(5)

375

except KeyboardInterrupt:

376

# cleanup

377

logging.config.stopListening()

378

t.join()

379

</pre></div>

380

</div>

381

And here is a script that takes a filename and sends that file to the server,

382

properly preceded with the binary-encoded length, as the new logging

383

configuration:

384

<div class="highlight-python3"><div class="highlight"><pre>#!/usr/bin/env python

385

import socket, sys, struct

386

387

with open(sys.argv[1], 'rb') as f:

388

data_to_send = f.read()

389

390

HOST = 'localhost'

391

PORT = 9999

392

s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)

393

print('connecting...')

394

s.connect((HOST, PORT))

395

print('sending config...')

396

s.send(struct.pack('>L', len(data_to_send)))

397

s.send(data_to_send)

398

s.close()

399

print('complete')

400

</pre></div>

401

</div>

402

</div>

403

404

<h2>Dealing with handlers that block<a class="headerlink" href="#dealing-with-handlers-that-block" title="Permalink to this headline">¶</a></h2>

405

Sometimes you have to get your logging handlers to do their work without

406

blocking the thread you’re logging from. This is common in Web applications,

407

though of course it also occurs in other scenarios.

408

A common culprit which demonstrates sluggish behaviour is the

409

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.SMTPHandler" title="logging.handlers.SMTPHandler"><code class="xref py py-class docutils literal">SMTPHandler</code></a>: sending emails can take a long time, for a

410

number of reasons outside the developer’s control (for example, a poorly

411

performing mail or network infrastructure). But almost any network-based

412

handler can block: Even a <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.SocketHandler" title="logging.handlers.SocketHandler"><code class="xref py py-class docutils literal">SocketHandler</code></a> operation may do a

413

DNS query under the hood which is too slow (and this query can be deep in the

414

socket library code, below the Python layer, and outside your control).

415

One solution is to use a two-part approach. For the first part, attach only a

416

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueHandler" title="logging.handlers.QueueHandler"><code class="xref py py-class docutils literal">QueueHandler</code></a> to those loggers which are accessed from

417

performance-critical threads. They simply write to their queue, which can be

418

sized to a large enough capacity or initialized with no upper bound to their

419

size. The write to the queue will typically be accepted quickly, though you

420

will probably need to catch the <a class="reference internal" href="../library/queue.html#queue.Full" title="queue.Full"><code class="xref py py-exc docutils literal">queue.Full</code></a> exception as a precaution

421

in your code. If you are a library developer who has performance-critical

422

threads in their code, be sure to document this (together with a suggestion to

423

attach only <code class="docutils literal">QueueHandlers</code> to your loggers) for the benefit of other

424

developers who will use your code.

425

The second part of the solution is <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueListener" title="logging.handlers.QueueListener"><code class="xref py py-class docutils literal">QueueListener</code></a>, which has been

426

designed as the counterpart to <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueHandler" title="logging.handlers.QueueHandler"><code class="xref py py-class docutils literal">QueueHandler</code></a>. A

427

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueListener" title="logging.handlers.QueueListener"><code class="xref py py-class docutils literal">QueueListener</code></a> is very simple: it’s passed a queue and some handlers,

428

and it fires up an internal thread which listens to its queue for LogRecords

429

sent from <code class="docutils literal">QueueHandlers</code> (or any other source of <code class="docutils literal">LogRecords</code>, for that

430

matter). The <code class="docutils literal">LogRecords</code> are removed from the queue and passed to the

431

handlers for processing.

432

The advantage of having a separate <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueListener" title="logging.handlers.QueueListener"><code class="xref py py-class docutils literal">QueueListener</code></a> class is that you

433

can use the same instance to service multiple <code class="docutils literal">QueueHandlers</code>. This is more

434

resource-friendly than, say, having threaded versions of the existing handler

435

classes, which would eat up one thread per handler for no particular benefit.

436

An example of using these two classes follows (imports omitted):

437

<div class="highlight-python3"><div class="highlight"><pre>que = queue.Queue(-1) # no limit on size

438

queue_handler = QueueHandler(que)

439

handler = logging.StreamHandler()

440

listener = QueueListener(que, handler)

441

root = logging.getLogger()

442

root.addHandler(queue_handler)

443

formatter = logging.Formatter('%(threadName)s: %(message)s')

444

handler.setFormatter(formatter)

445

listener.start()

446

# The log output will display the thread which generated

447

# the event (the main thread) rather than the internal

448

# thread which monitors the internal queue. This is what

449

# you want to happen.

450

root.warning('Look out!')

451

listener.stop()

452

</pre></div>

453

</div>

454

which, when run, will produce:

455

<div class="highlight-none"><div class="highlight"><pre>MainThread: Look out!

456

</pre></div>

457

</div>

458

459

Changed in version 3.5: Prior to Python 3.5, the <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueListener" title="logging.handlers.QueueListener"><code class="xref py py-class docutils literal">QueueListener</code></a> always passed every message

460

received from the queue to every handler it was initialized with. (This was

461

because it was assumed that level filtering was all done on the other side,

462

where the queue is filled.) From 3.5 onwards, this behaviour can be changed

463

by passing a keyword argument <code class="docutils literal">respect_handler_level=True</code> to the

464

listener’s constructor. When this is done, the listener compares the level

465

of each message with the handler’s level, and only passes a message to a

466

handler if it’s appropriate to do so.

467

</div>

468

</div>

469

470

<h2>Sending and receiving logging events across a network<a class="headerlink" href="#sending-and-receiving-logging-events-across-a-network" title="Permalink to this headline">¶</a></h2>

471

Let’s say you want to send logging events across a network, and handle them at

472

the receiving end. A simple way of doing this is attaching a

473

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.SocketHandler" title="logging.handlers.SocketHandler"><code class="xref py py-class docutils literal">SocketHandler</code></a> instance to the root logger at the sending end:

474

<div class="highlight-python3"><div class="highlight"><pre>import logging, logging.handlers

475

476

rootLogger = logging.getLogger('')

477

rootLogger.setLevel(logging.DEBUG)

478

socketHandler = logging.handlers.SocketHandler('localhost',

479

logging.handlers.DEFAULT_TCP_LOGGING_PORT)

480

# don't bother with a formatter, since a socket handler sends the event as

481

# an unformatted pickle

482

rootLogger.addHandler(socketHandler)

483

484

# Now, we can log to the root logger, or any other logger. First the root...

485

logging.info('Jackdaws love my big sphinx of quartz.')

486

487

# Now, define a couple of other loggers which might represent areas in your

488

# application:

489

490

491

492

493

logger1.debug('Quick zephyrs blow, vexing daft Jim.')

494

logger1.info('How quickly daft jumping zebras vex.')

495

496

logger2.error('The five boxing wizards jump quickly.')

497

</pre></div>

498

</div>

499

At the receiving end, you can set up a receiver using the <a class="reference internal" href="../library/socketserver.html#module-socketserver" title="socketserver: A framework for network servers."><code class="xref py py-mod docutils literal">socketserver</code></a>

500

module. Here is a basic working example:

501

<div class="highlight-python3"><div class="highlight"><pre>import pickle

502

import logging

503

import logging.handlers

504

import socketserver

505

import struct

506

507

508

class LogRecordStreamHandler(socketserver.StreamRequestHandler):

509

"""Handler for a streaming logging request.

510

511

This basically logs the record using whatever logging policy is

512

configured locally.

513

"""

514

515

def handle(self):

516

"""

517

Handle multiple requests - each expected to be a 4-byte length,

518

followed by the LogRecord in pickle format. Logs the record

519

according to whatever policy is configured locally.

520

"""

521

while True:

522

chunk = self.connection.recv(4)

523

if len(chunk) < 4:

524

break

525

slen = struct.unpack('>L', chunk)[0]

526

527

while len(chunk) < slen:

528

chunk = chunk + self.connection.recv(slen - len(chunk))

529

obj = self.unPickle(chunk)

530

record = logging.makeLogRecord(obj)

531

self.handleLogRecord(record)

532

533

def unPickle(self, data):

534

return pickle.loads(data)

535

536

def handleLogRecord(self, record):

537

# if a name is specified, we use the named logger rather than the one

538

# implied by the record.

539

if self.server.logname is not None:

540

name = self.server.logname

541

else:

542

name = record.name

543

544

# N.B. EVERY record gets logged. This is because Logger.handle

545

# is normally called AFTER logger-level filtering. If you want

546

# to do filtering, do it at the client end to save wasting

547

# cycles and network bandwidth!

548

logger.handle(record)

549

550

class LogRecordSocketReceiver(socketserver.ThreadingTCPServer):

551

"""

552

Simple TCP socket-based logging receiver suitable for testing.

553

"""

554

555

allow_reuse_address = True

556

557

def __init__(self, host='localhost',

558

port=logging.handlers.DEFAULT_TCP_LOGGING_PORT,

559

handler=LogRecordStreamHandler):

560

socketserver.ThreadingTCPServer.__init__(self, (host, port), handler)

561

self.abort = 0

562

self.timeout = 1

563

self.logname = None

564

565

def serve_until_stopped(self):

566

import select

567

abort = 0

568

while not abort:

569

rd, wr, ex = select.select([self.socket.fileno()],

570

[], [],

571

self.timeout)

572

if rd:

573

self.handle_request()

574

abort = self.abort

575

576

def main():

577

logging.basicConfig(

578

format='%(relativeCreated)5d %(name)-15s %(levelname)-8s %(message)s')

579

tcpserver = LogRecordSocketReceiver()

580

print('About to start TCP server...')

581

tcpserver.serve_until_stopped()

582

583

if __name__ == '__main__':

584

main()

585

</pre></div>

586

</div>

587

First run the server, and then the client. On the client side, nothing is

588

printed on the console; on the server side, you should see something like:

589

<div class="highlight-python3"><div class="highlight"><pre>About to start TCP server...

590

59 root INFO Jackdaws love my big sphinx of quartz.

591

59 myapp.area1 DEBUG Quick zephyrs blow, vexing daft Jim.

592

69 myapp.area1 INFO How quickly daft jumping zebras vex.

593

69 myapp.area2 WARNING Jail zesty vixen who grabbed pay from quack.

594

69 myapp.area2 ERROR The five boxing wizards jump quickly.

595

</pre></div>

596

</div>

597

Note that there are some security issues with pickle in some scenarios. If

598

these affect you, you can use an alternative serialization scheme by overriding

599

the <code class="xref py py-meth docutils literal">makePickle()</code> method and implementing your

600

alternative there, as well as adapting the above script to use your alternative

601

serialization.

602

</div>

603

604

<h2>Adding contextual information to your logging output<a class="headerlink" href="#adding-contextual-information-to-your-logging-output" title="Permalink to this headline">¶</a></h2>

605

Sometimes you want logging output to contain contextual information in

606

addition to the parameters passed to the logging call. For example, in a

607

networked application, it may be desirable to log client-specific information

608

in the log (e.g. remote client’s username, or IP address). Although you could

609

use the extra parameter to achieve this, it’s not always convenient to pass

610

the information in this way. While it might be tempting to create

611

<code class="xref py py-class docutils literal">Logger</code> instances on a per-connection basis, this is not a good idea

612

because these instances are not garbage collected. While this is not a problem

613

in practice, when the number of <code class="xref py py-class docutils literal">Logger</code> instances is dependent on the

614

level of granularity you want to use in logging an application, it could

615

be hard to manage if the number of <code class="xref py py-class docutils literal">Logger</code> instances becomes

616

effectively unbounded.

617

618

<h3>Using LoggerAdapters to impart contextual information<a class="headerlink" href="#using-loggeradapters-to-impart-contextual-information" title="Permalink to this headline">¶</a></h3>

619

An easy way in which you can pass contextual information to be output along

620

with logging event information is to use the <code class="xref py py-class docutils literal">LoggerAdapter</code> class.

621

This class is designed to look like a <code class="xref py py-class docutils literal">Logger</code>, so that you can call

622

<code class="xref py py-meth docutils literal">debug()</code>, <code class="xref py py-meth docutils literal">info()</code>, <code class="xref py py-meth docutils literal">warning()</code>, <code class="xref py py-meth docutils literal">error()</code>,

623

<code class="xref py py-meth docutils literal">exception()</code>, <code class="xref py py-meth docutils literal">critical()</code> and <code class="xref py py-meth docutils literal">log()</code>. These methods have the

624

same signatures as their counterparts in <code class="xref py py-class docutils literal">Logger</code>, so you can use the

625

two types of instances interchangeably.

626

When you create an instance of <code class="xref py py-class docutils literal">LoggerAdapter</code>, you pass it a

627

<code class="xref py py-class docutils literal">Logger</code> instance and a dict-like object which contains your contextual

628

information. When you call one of the logging methods on an instance of

629

<code class="xref py py-class docutils literal">LoggerAdapter</code>, it delegates the call to the underlying instance of

630

<code class="xref py py-class docutils literal">Logger</code> passed to its constructor, and arranges to pass the contextual

631

information in the delegated call. Here’s a snippet from the code of

632

<code class="xref py py-class docutils literal">LoggerAdapter</code>:

633

<div class="highlight-python3"><div class="highlight"><pre>def debug(self, msg, *args, **kwargs):

634

"""

635

Delegate a debug call to the underlying logger, after adding

636

contextual information from this adapter instance.

637

"""

638

msg, kwargs = self.process(msg, kwargs)

639

self.logger.debug(msg, *args, **kwargs)

640

</pre></div>

641

</div>

642

The <code class="xref py py-meth docutils literal">process()</code> method of <code class="xref py py-class docutils literal">LoggerAdapter</code> is where the

643

contextual information is added to the logging output. It’s passed the message

644

and keyword arguments of the logging call, and it passes back (potentially)

645

modified versions of these to use in the call to the underlying logger. The

646

default implementation of this method leaves the message alone, but inserts

647

an ‘extra’ key in the keyword argument whose value is the dict-like object

648

passed to the constructor. Of course, if you had passed an ‘extra’ keyword

649

argument in the call to the adapter, it will be silently overwritten.

650

The advantage of using ‘extra’ is that the values in the dict-like object are

651

merged into the <code class="xref py py-class docutils literal">LogRecord</code> instance’s __dict__, allowing you to use

652

customized strings with your <code class="xref py py-class docutils literal">Formatter</code> instances which know about

653

the keys of the dict-like object. If you need a different method, e.g. if you

654

want to prepend or append the contextual information to the message string,

655

you just need to subclass <code class="xref py py-class docutils literal">LoggerAdapter</code> and override

656

<code class="xref py py-meth docutils literal">process()</code> to do what you need. Here is a simple example:

657

<div class="highlight-python3"><div class="highlight"><pre>class CustomAdapter(logging.LoggerAdapter):

658

"""

659

This example adapter expects the passed in dict-like object to have a

660

'connid' key, whose value in brackets is prepended to the log message.

661

"""

662

def process(self, msg, kwargs):

663

return '[%s] %s' % (self.extra['connid'], msg), kwargs

664

</pre></div>

665

</div>

666

which you can use like this:

667

<div class="highlight-python3"><div class="highlight"><pre>logger = logging.getLogger(__name__)

668

adapter = CustomAdapter(logger, {'connid': some_conn_id})

669

</pre></div>

670

</div>

671

Then any events that you log to the adapter will have the value of

672

<code class="docutils literal">some_conn_id</code> prepended to the log messages.

673

674

<h4>Using objects other than dicts to pass contextual information<a class="headerlink" href="#using-objects-other-than-dicts-to-pass-contextual-information" title="Permalink to this headline">¶</a></h4>

675

You don’t need to pass an actual dict to a <code class="xref py py-class docutils literal">LoggerAdapter</code> - you could

676

pass an instance of a class which implements <code class="docutils literal">__getitem__</code> and <code class="docutils literal">__iter__</code> so

677

that it looks like a dict to logging. This would be useful if you want to

678

generate values dynamically (whereas the values in a dict would be constant).

679

</div>

680

</div>

681

682

<h3>Using Filters to impart contextual information<a class="headerlink" href="#using-filters-to-impart-contextual-information" title="Permalink to this headline">¶</a></h3>

683

You can also add contextual information to log output using a user-defined

684

<code class="xref py py-class docutils literal">Filter</code>. <code class="docutils literal">Filter</code> instances are allowed to modify the <code class="docutils literal">LogRecords</code>

685

passed to them, including adding additional attributes which can then be output

686

using a suitable format string, or if needed a custom <code class="xref py py-class docutils literal">Formatter</code>.

687

For example in a web application, the request being processed (or at least,

688

the interesting parts of it) can be stored in a threadlocal

689

(<a class="reference internal" href="../library/threading.html#threading.local" title="threading.local"><code class="xref py py-class docutils literal">threading.local</code></a>) variable, and then accessed from a <code class="docutils literal">Filter</code> to

690

add, say, information from the request - say, the remote IP address and remote

691

user’s username - to the <code class="docutils literal">LogRecord</code>, using the attribute names ‘ip’ and

692

‘user’ as in the <code class="docutils literal">LoggerAdapter</code> example above. In that case, the same format

693

string can be used to get similar output to that shown above. Here’s an example

694

script:

695

<div class="highlight-python3"><div class="highlight"><pre>import logging

696

from random import choice

697

698

class ContextFilter(logging.Filter):

699

"""

700

This is a filter which injects contextual information into the log.

701

702

Rather than use actual contextual information, we just use random

703

data in this demo.

704

"""

705

706

USERS = ['jim', 'fred', 'sheila']

707

IPS = ['123.231.231.123', '127.0.0.1', '192.168.0.1']

708

709

def filter(self, record):

710

711

record.ip = choice(ContextFilter.IPS)

712

record.user = choice(ContextFilter.USERS)

713

return True

714

715

if __name__ == '__main__':

716

levels = (logging.DEBUG, logging.INFO, logging.WARNING, logging.ERROR, logging.CRITICAL)

717

718

format='%(asctime)-15s %(name)-5s %(levelname)-8s IP: %(ip)-15s User: %(user)-8s %(message)s')

719

a1 = logging.getLogger('a.b.c')

720

a2 = logging.getLogger('d.e.f')

721

722

f = ContextFilter()

723

a1.addFilter(f)

724

a2.addFilter(f)

725

a1.debug('A debug message')

726

a1.info('An info message with %s', 'some parameters')

727

for x in range(10):

728

lvl = choice(levels)

729

lvlname = logging.getLevelName(lvl)

730

a2.log(lvl, 'A message at %s level with %d %s', lvlname, 2, 'parameters')

731

</pre></div>

732

</div>

733

which, when run, produces something like:

734

<div class="highlight-python3"><div class="highlight"><pre>2010-09-06 22:38:15,292 a.b.c DEBUG IP: 123.231.231.123 User: fred A debug message

735

2010-09-06 22:38:15,300 a.b.c INFO IP: 192.168.0.1 User: sheila An info message with some parameters

736

2010-09-06 22:38:15,300 d.e.f CRITICAL IP: 127.0.0.1 User: sheila A message at CRITICAL level with 2 parameters

737

2010-09-06 22:38:15,300 d.e.f ERROR IP: 127.0.0.1 User: jim A message at ERROR level with 2 parameters

738

2010-09-06 22:38:15,300 d.e.f DEBUG IP: 127.0.0.1 User: sheila A message at DEBUG level with 2 parameters

739

2010-09-06 22:38:15,300 d.e.f ERROR IP: 123.231.231.123 User: fred A message at ERROR level with 2 parameters

740

2010-09-06 22:38:15,300 d.e.f CRITICAL IP: 192.168.0.1 User: jim A message at CRITICAL level with 2 parameters

741

2010-09-06 22:38:15,300 d.e.f CRITICAL IP: 127.0.0.1 User: sheila A message at CRITICAL level with 2 parameters

742

2010-09-06 22:38:15,300 d.e.f DEBUG IP: 192.168.0.1 User: jim A message at DEBUG level with 2 parameters

743

2010-09-06 22:38:15,301 d.e.f ERROR IP: 127.0.0.1 User: sheila A message at ERROR level with 2 parameters

744

2010-09-06 22:38:15,301 d.e.f DEBUG IP: 123.231.231.123 User: fred A message at DEBUG level with 2 parameters

745

2010-09-06 22:38:15,301 d.e.f INFO IP: 123.231.231.123 User: fred A message at INFO level with 2 parameters

746

</pre></div>

747

</div>

748

</div>

749

</div>

750

751

<h2>Logging to a single file from multiple processes<a class="headerlink" href="#logging-to-a-single-file-from-multiple-processes" title="Permalink to this headline">¶</a></h2>

752

Although logging is thread-safe, and logging to a single file from multiple

753

threads in a single process is supported, logging to a single file from

754

multiple processes is not supported, because there is no standard way to

755

serialize access to a single file across multiple processes in Python. If you

756

need to log to a single file from multiple processes, one way of doing this is

757

to have all the processes log to a <code class="xref py py-class docutils literal">SocketHandler</code>, and have a

758

separate process which implements a socket server which reads from the socket

759

and logs to file. (If you prefer, you can dedicate one thread in one of the

760

existing processes to perform this function.)

761

<a class="reference internal" href="#network-logging">This section</a> documents this approach in more detail and

762

includes a working socket receiver which can be used as a starting point for you

763

to adapt in your own applications.

764

If you are using a recent version of Python which includes the

765

<a class="reference internal" href="../library/multiprocessing.html#module-multiprocessing" title="multiprocessing: Process-based parallelism."><code class="xref py py-mod docutils literal">multiprocessing</code></a> module, you could write your own handler which uses the

766

<a class="reference internal" href="../library/multiprocessing.html#multiprocessing.Lock" title="multiprocessing.Lock"><code class="xref py py-class docutils literal">Lock</code></a> class from this module to serialize access to the

767

file from your processes. The existing <code class="xref py py-class docutils literal">FileHandler</code> and subclasses do

768

not make use of <a class="reference internal" href="../library/multiprocessing.html#module-multiprocessing" title="multiprocessing: Process-based parallelism."><code class="xref py py-mod docutils literal">multiprocessing</code></a> at present, though they may do so in the

769

future. Note that at present, the <a class="reference internal" href="../library/multiprocessing.html#module-multiprocessing" title="multiprocessing: Process-based parallelism."><code class="xref py py-mod docutils literal">multiprocessing</code></a> module does not provide

770

working lock functionality on all platforms (see

771

<a class="reference external" href="https://bugs.python.org/issue3770">https://bugs.python.org/issue3770</a>).

772

Alternatively, you can use a <code class="docutils literal">Queue</code> and a <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.QueueHandler" title="logging.handlers.QueueHandler"><code class="xref py py-class docutils literal">QueueHandler</code></a> to send

773

all logging events to one of the processes in your multi-process application.

774

The following example script demonstrates how you can do this; in the example

775

a separate listener process listens for events sent by other processes and logs

776

them according to its own logging configuration. Although the example only

777

demonstrates one way of doing it (for example, you may want to use a listener

778

thread rather than a separate listener process – the implementation would be

779

analogous) it does allow for completely different logging configurations for

780

the listener and the other processes in your application, and can be used as

781

the basis for code meeting your own specific requirements:

782

<div class="highlight-python3"><div class="highlight"><pre># You'll need these imports in your own code

783

import logging

784

import logging.handlers

785

import multiprocessing

786

787

# Next two import lines for this demo only

788

from random import choice, random

789

import time

790

791

#

792

# Because you'll want to define the logging configurations for listener and workers, the

793

# listener and worker process functions take a configurer parameter which is a callable

794

# for configuring logging for that process. These functions are also passed the queue,

795

# which they use for communication.

796

#

797

# In practice, you can configure the listener however you want, but note that in this

798

# simple example, the listener does not apply level or filter logic to received records.

799

# In practice, you would probably want to do this logic in the worker processes, to avoid

800

# sending events which would be filtered out between processes.

801

#

802

# The size of the rotated files is made small so you can see the results easily.

803

def listener_configurer():

804

root = logging.getLogger()

805

h = logging.handlers.RotatingFileHandler('mptest.log', 'a', 300, 10)

806

f = logging.Formatter('%(asctime)s %(processName)-10s %(name)s %(levelname)-8s %(message)s')

807

h.setFormatter(f)

808

root.addHandler(h)

809

810

# This is the listener process top-level loop: wait for logging events

811

# (LogRecords)on the queue and handle them, quit when you get a None for a

812

# LogRecord.

813

def listener_process(queue, configurer):

814

configurer()

815

while True:

816

try:

817

record = queue.get()

818

if record is None: # We send this as a sentinel to tell the listener to quit.

819

break

820

logger = logging.getLogger(record.name)

821

logger.handle(record) # No level or filter logic applied - just do it!

822

except Exception:

823

import sys, traceback

824

print('Whoops! Problem:', file=sys.stderr)

825

traceback.print_exc(file=sys.stderr)

826

827

# Arrays used for random selections in this demo

828

829

LEVELS = [logging.DEBUG, logging.INFO, logging.WARNING,

830

logging.ERROR, logging.CRITICAL]

831

832

LOGGERS = ['a.b.c', 'd.e.f']

833

834

MESSAGES = [

835

'Random message #1',

836

'Random message #2',

837

'Random message #3',

838

]

839

840

# The worker configuration is done at the start of the worker process run.

841

# Note that on Windows you can't rely on fork semantics, so each process

842

# will run the logging configuration code when it starts.

843

def worker_configurer(queue):

844

h = logging.handlers.QueueHandler(queue) # Just the one handler needed

845

root = logging.getLogger()

846

root.addHandler(h)

847

# send all messages, for demo; no other level or filter logic applied.

848

root.setLevel(logging.DEBUG)

849

850

# This is the worker process top-level loop, which just logs ten events with

851

# random intervening delays before terminating.

852

# The print messages are just so you know it's doing something!

853

def worker_process(queue, configurer):

854

configurer(queue)

855

name = multiprocessing.current_process().name

856

print('Worker started: %s' % name)

857

for i in range(10):

858

time.sleep(random())

859

logger = logging.getLogger(choice(LOGGERS))

860

level = choice(LEVELS)

861

message = choice(MESSAGES)

862

logger.log(level, message)

863

print('Worker finished: %s' % name)

864

865

# Here's where the demo gets orchestrated. Create the queue, create and start

866

# the listener, create ten workers and start them, wait for them to finish,

867

# then send a None to the queue to tell the listener to finish.

868

def main():

869

queue = multiprocessing.Queue(-1)

870

listener = multiprocessing.Process(target=listener_process,

871

args=(queue, listener_configurer))

872

listener.start()

873

workers = []

874

for i in range(10):

875

worker = multiprocessing.Process(target=worker_process,

876

args=(queue, worker_configurer))

877

workers.append(worker)

878

worker.start()

879

for w in workers:

880

w.join()

881

queue.put_nowait(None)

882

listener.join()

883

884

if __name__ == '__main__':

885

main()

886

</pre></div>

887

</div>

888

A variant of the above script keeps the logging in the main process, in a

889

separate thread:

890

<div class="highlight-python3"><div class="highlight"><pre>import logging

891

import logging.config

892

import logging.handlers

893

from multiprocessing import Process, Queue

894

import random

895

import threading

896

import time

897

898

def logger_thread(q):

899

while True:

900

record = q.get()

901

if record is None:

902

break

903

logger = logging.getLogger(record.name)

904

logger.handle(record)

905

906

907

def worker_process(q):

908

qh = logging.handlers.QueueHandler(q)

909

root = logging.getLogger()

910

911

root.addHandler(qh)

912

levels = [logging.DEBUG, logging.INFO, logging.WARNING, logging.ERROR,

913

logging.CRITICAL]

914

loggers = ['foo', 'foo.bar', 'foo.bar.baz',

915

'spam', 'spam.ham', 'spam.ham.eggs']

916

for i in range(100):

917

lvl = random.choice(levels)

918

logger = logging.getLogger(random.choice(loggers))

919

logger.log(lvl, 'Message no. %d', i)

920

921

if __name__ == '__main__':

922

q = Queue()

923

d = {

924

'version': 1,

925

'formatters': {

926

'detailed': {

927

'class': 'logging.Formatter',

928

'format': '%(asctime)s %(name)-15s %(levelname)-8s %(processName)-10s %(message)s'

929

}

930

},

931

'handlers': {

932

'console': {

933

'class': 'logging.StreamHandler',

934

'level': 'INFO',

935

},

936

'file': {

937

'class': 'logging.FileHandler',

938

'filename': 'mplog.log',

939

'mode': 'w',

940

'formatter': 'detailed',

941

},

942

'foofile': {

943

'class': 'logging.FileHandler',

944

'filename': 'mplog-foo.log',

945

'mode': 'w',

946

'formatter': 'detailed',

947

},

948

'errors': {

949

'class': 'logging.FileHandler',

950

'filename': 'mplog-errors.log',

951

'mode': 'w',

952

'level': 'ERROR',

953

'formatter': 'detailed',

954

},

955

},

956

'loggers': {

957

'foo': {

958

'handlers': ['foofile']

959

}

960

},

961

'root': {

962

'level': 'DEBUG',

963

'handlers': ['console', 'file', 'errors']

964

},

965

}

966

workers = []

967

for i in range(5):

968

wp = Process(target=worker_process, name='worker %d' % (i + 1), args=(q,))

969

workers.append(wp)

970

wp.start()

971

logging.config.dictConfig(d)

972

lp = threading.Thread(target=logger_thread, args=(q,))

973

lp.start()

974

# At this point, the main process could do some useful work of its own

975

# Once it's done that, it can wait for the workers to terminate...

976

for wp in workers:

977

wp.join()

978

# And now tell the logging thread to finish up, too

979

q.put(None)

980

lp.join()

981

</pre></div>

982

</div>

983

This variant shows how you can e.g. apply configuration for particular loggers

984

- e.g. the <code class="docutils literal">foo</code> logger has a special handler which stores all events in the

985

<code class="docutils literal">foo</code> subsystem in a file <code class="docutils literal">mplog-foo.log</code>. This will be used by the logging

986

machinery in the main process (even though the logging events are generated in

987

the worker processes) to direct the messages to the appropriate destinations.

988

</div>

989

990

<h2>Using file rotation<a class="headerlink" href="#using-file-rotation" title="Permalink to this headline">¶</a></h2>

991

Sometimes you want to let a log file grow to a certain size, then open a new

992

file and log to that. You may want to keep a certain number of these files, and

993

when that many files have been created, rotate the files so that the number of

994

files and the size of the files both remain bounded. For this usage pattern, the

995

logging package provides a <code class="xref py py-class docutils literal">RotatingFileHandler</code>:

996

<div class="highlight-python3"><div class="highlight"><pre>import glob

997

import logging

998

import logging.handlers

999

1000

LOG_FILENAME = 'logging_rotatingfile_example.out'

1001

1002

# Set up a specific logger with our desired output level

1003

my_logger = logging.getLogger('MyLogger')

1004

my_logger.setLevel(logging.DEBUG)

1005

1006

# Add the log message handler to the logger

1007

handler = logging.handlers.RotatingFileHandler(

1008

LOG_FILENAME, maxBytes=20, backupCount=5)

1009

1010

my_logger.addHandler(handler)

1011

1012

# Log some messages

1013

for i in range(20):

1014

my_logger.debug('i = %d' % i)

1015

1016

# See what files are created

1017

logfiles = glob.glob('%s*' % LOG_FILENAME)

1018

1019

for filename in logfiles:

1020

print(filename)

1021

</pre></div>

1022

</div>

1023

The result should be 6 separate files, each with part of the log history for the

1024

application:

1025

<div class="highlight-python3"><div class="highlight"><pre>logging_rotatingfile_example.out

1026

logging_rotatingfile_example.out.1

1027

logging_rotatingfile_example.out.2

1028

logging_rotatingfile_example.out.3

1029

logging_rotatingfile_example.out.4

1030

logging_rotatingfile_example.out.5

1031

</pre></div>

1032

</div>

1033

The most current file is always <code class="file docutils literal">logging_rotatingfile_example.out</code>,

1034

and each time it reaches the size limit it is renamed with the suffix

1035

<code class="docutils literal">.1</code>. Each of the existing backup files is renamed to increment the suffix

1036

(<code class="docutils literal">.1</code> becomes <code class="docutils literal">.2</code>, etc.) and the <code class="docutils literal">.6</code> file is erased.

1037

Obviously this example sets the log length much too small as an extreme

1038

example. You would want to set maxBytes to an appropriate value.

1039

</div>

1040

1041

<h2>Use of alternative formatting styles<a class="headerlink" href="#use-of-alternative-formatting-styles" title="Permalink to this headline">¶</a></h2>

1042

When logging was added to the Python standard library, the only way of

1043

formatting messages with variable content was to use the %-formatting

1044

method. Since then, Python has gained two new formatting approaches:

1045

<a class="reference internal" href="../library/string.html#string.Template" title="string.Template"><code class="xref py py-class docutils literal">string.Template</code></a> (added in Python 2.4) and <a class="reference internal" href="../library/stdtypes.html#str.format" title="str.format"><code class="xref py py-meth docutils literal">str.format()</code></a>

1046

(added in Python 2.6).

1047

Logging (as of 3.2) provides improved support for these two additional

1048

formatting styles. The <code class="xref py py-class docutils literal">Formatter</code> class been enhanced to take an

1049

additional, optional keyword parameter named <code class="docutils literal">style</code>. This defaults to

1050

<code class="docutils literal">'%'</code>, but other possible values are <code class="docutils literal">'{'</code> and <code class="docutils literal">'$'</code>, which correspond

1051

to the other two formatting styles. Backwards compatibility is maintained by

1052

default (as you would expect), but by explicitly specifying a style parameter,

1053

you get the ability to specify format strings which work with

1054

<a class="reference internal" href="../library/stdtypes.html#str.format" title="str.format"><code class="xref py py-meth docutils literal">str.format()</code></a> or <a class="reference internal" href="../library/string.html#string.Template" title="string.Template"><code class="xref py py-class docutils literal">string.Template</code></a>. Here’s an example console

1055

session to show the possibilities:

1056

<div class="highlight-pycon"><div class="highlight"><pre>>>> import logging

1057

>>> root = logging.getLogger()

1058

>>> root.setLevel(logging.DEBUG)

1059

>>> handler = logging.StreamHandler()

1060

>>> bf = logging.Formatter('{asctime} {name} {levelname:8s} {message}',

1061

... style='{')

1062

>>> handler.setFormatter(bf)

1063

>>> root.addHandler(handler)

1064

>>> logger = logging.getLogger('foo.bar')

1065

>>> logger.debug('This is a DEBUG message')

1066

2010-10-28 15:11:55,341 foo.bar DEBUG This is a DEBUG message

1067

>>> logger.critical('This is a CRITICAL message')

1068

2010-10-28 15:12:11,526 foo.bar CRITICAL This is a CRITICAL message

1069

>>> df = logging.Formatter('$asctime $name ${levelname} $message',

1070

... style='$')

1071

>>> handler.setFormatter(df)

1072

1073

2010-10-28 15:13:06,924 foo.bar DEBUG This is a DEBUG message

1074

1075

2010-10-28 15:13:11,494 foo.bar CRITICAL This is a CRITICAL message

1076

>>>

1077

</pre></div>

1078

</div>

1079

Note that the formatting of logging messages for final output to logs is

1080

completely independent of how an individual logging message is constructed.

1081

That can still use %-formatting, as shown here:

1082

<div class="highlight-python3"><div class="highlight"><pre>>>> logger.error('This is an%s %s %s', 'other,', 'ERROR,', 'message')

1083

2010-10-28 15:19:29,833 foo.bar ERROR This is another, ERROR, message

1084

>>>

1085

</pre></div>

1086

</div>

1087

Logging calls (<code class="docutils literal">logger.debug()</code>, <code class="docutils literal">logger.info()</code> etc.) only take

1088

positional parameters for the actual logging message itself, with keyword

1089

parameters used only for determining options for how to handle the actual

1090

logging call (e.g. the <code class="docutils literal">exc_info</code> keyword parameter to indicate that

1091

traceback information should be logged, or the <code class="docutils literal">extra</code> keyword parameter

1092

to indicate additional contextual information to be added to the log). So

1093

you cannot directly make logging calls using <a class="reference internal" href="../library/stdtypes.html#str.format" title="str.format"><code class="xref py py-meth docutils literal">str.format()</code></a> or

1094

1095

uses %-formatting to merge the format string and the variable arguments.

1096

There would no changing this while preserving backward compatibility, since

1097

all logging calls which are out there in existing code will be using %-format

1098

strings.

1099

There is, however, a way that you can use {}- and $- formatting to construct

1100

your individual log messages. Recall that for a message you can use an

1101

arbitrary object as a message format string, and that the logging package will

1102

call <code class="docutils literal">str()</code> on that object to get the actual format string. Consider the

1103

following two classes:

1104

<div class="highlight-python3"><div class="highlight"><pre>class BraceMessage:

1105

def __init__(self, fmt, *args, **kwargs):

1106

self.fmt = fmt

1107

self.args = args

1108

self.kwargs = kwargs

1109

1110

def __str__(self):

1111

return self.fmt.format(*self.args, **self.kwargs)

1112

1113

class DollarMessage:

1114

def __init__(self, fmt, **kwargs):

1115

self.fmt = fmt

1116

self.kwargs = kwargs

1117

1118

def __str__(self):

1119

from string import Template

1120

return Template(self.fmt).substitute(**self.kwargs)

1121

</pre></div>

1122

</div>

1123

Either of these can be used in place of a format string, to allow {}- or

1124

$-formatting to be used to build the actual “message” part which appears in the

1125

formatted log output in place of “%(message)s” or “{message}” or “$message”.

1126

It’s a little unwieldy to use the class names whenever you want to log

1127

something, but it’s quite palatable if you use an alias such as __ (double

1128

underscore – not to be confused with _, the single underscore used as a

1129

synonym/alias for <a class="reference internal" href="../library/gettext.html#gettext.gettext" title="gettext.gettext"><code class="xref py py-func docutils literal">gettext.gettext()</code></a> or its brethren).

1130

The above classes are not included in Python, though they’re easy enough to

1131

copy and paste into your own code. They can be used as follows (assuming that

1132

they’re declared in a module called <code class="docutils literal">wherever</code>):

1133

<div class="highlight-pycon"><div class="highlight"><pre>>>> from wherever import BraceMessage as __

1134

>>> print(__('Message with {0} {name}', 2, name='placeholders'))

1135

Message with 2 placeholders

1136

>>> class Point: pass

1137

...

1138

>>> p = Point()

1139

>>> p.x = 0.5

1140

>>> p.y = 0.5

1141

1142

... point=p))

1143

Message with coordinates: (0.50, 0.50)

1144

>>> from wherever import DollarMessage as __

1145

>>> print(__('Message with $num $what', num=2, what='placeholders'))

1146

Message with 2 placeholders

1147

>>>

1148

</pre></div>

1149

</div>

1150

While the above examples use <code class="docutils literal">print()</code> to show how the formatting works, you

1151

would of course use <code class="docutils literal">logger.debug()</code> or similar to actually log using this

1152

approach.

1153

One thing to note is that you pay no significant performance penalty with this

1154

approach: the actual formatting happens not when you make the logging call, but

1155

when (and if) the logged message is actually about to be output to a log by a

1156

handler. So the only slightly unusual thing which might trip you up is that the

1157

parentheses go around the format string and the arguments, not just the format

1158

string. That’s because the __ notation is just syntax sugar for a constructor

1159

call to one of the XXXMessage classes.

1160

If you prefer, you can use a <code class="xref py py-class docutils literal">LoggerAdapter</code> to achieve a similar effect

1161

to the above, as in the following example:

1162

<div class="highlight-python3"><div class="highlight"><pre>import logging

1163

1164

class Message(object):

1165

1166

self.fmt = fmt

1167

self.args = args

1168

1169

def __str__(self):

1170

1171

1172

class StyleAdapter(logging.LoggerAdapter):

1173

1174

super(StyleAdapter, self).__init__(logger, extra or {})

1175

1176

def log(self, level, msg, *args, **kwargs):

1177

if self.isEnabledFor(level):

1178

1179

self.logger._log(level, Message(msg, args), (), **kwargs)

1180

1181

logger = StyleAdapter(logging.getLogger(__name__))

1182

1183

def main():

1184

logger.debug('Hello, {}', 'world!')

1185

1186

if __name__ == '__main__':

1187

1188

main()

1189

</pre></div>

1190

</div>

1191

The above script should log the message <code class="docutils literal">Hello, world!</code> when run with

1192

Python 3.2 or later.

1193

</div>

1194

1195

<h2>Customizing <code class="docutils literal">LogRecord</code><a class="headerlink" href="#customizing-logrecord" title="Permalink to this headline">¶</a></h2>

1196

Every logging event is represented by a <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a> instance.

1197

When an event is logged and not filtered out by a logger’s level, a

1198

1199

then passed to the handlers for that logger (and its ancestors, up to and

1200

including the logger where further propagation up the hierarchy is disabled).

1201

Before Python 3.2, there were only two places where this creation was done:

1202

1203

<li><a class="reference internal" href="../library/logging.html#logging.Logger.makeRecord" title="logging.Logger.makeRecord"><code class="xref py py-meth docutils literal">Logger.makeRecord()</code></a>, which is called in the normal process of

1204

logging an event. This invoked <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a> directly to create an

1205

instance.</li>

1206

<li><a class="reference internal" href="../library/logging.html#logging.makeLogRecord" title="logging.makeLogRecord"><code class="xref py py-func docutils literal">makeLogRecord()</code></a>, which is called with a dictionary containing

1207

attributes to be added to the LogRecord. This is typically invoked when a

1208

suitable dictionary has been received over the network (e.g. in pickle form

1209

via a <a class="reference internal" href="../library/logging.handlers.html#logging.handlers.SocketHandler" title="logging.handlers.SocketHandler"><code class="xref py py-class docutils literal">SocketHandler</code></a>, or in JSON form via an

1210

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.HTTPHandler" title="logging.handlers.HTTPHandler"><code class="xref py py-class docutils literal">HTTPHandler</code></a>).</li>

1211

</ul>

1212

This has usually meant that if you need to do anything special with a

1213

1214

1215

<li>Create your own <a class="reference internal" href="../library/logging.html#logging.Logger" title="logging.Logger"><code class="xref py py-class docutils literal">Logger</code></a> subclass, which overrides

1216

<a class="reference internal" href="../library/logging.html#logging.Logger.makeRecord" title="logging.Logger.makeRecord"><code class="xref py py-meth docutils literal">Logger.makeRecord()</code></a>, and set it using <a class="reference internal" href="../library/logging.html#logging.setLoggerClass" title="logging.setLoggerClass"><code class="xref py py-func docutils literal">setLoggerClass()</code></a>

1217

before any loggers that you care about are instantiated.</li>

1218

<li>Add a <a class="reference internal" href="../library/logging.html#logging.Filter" title="logging.Filter"><code class="xref py py-class docutils literal">Filter</code></a> to a logger or handler, which does the

1219

necessary special manipulation you need when its

1220

<a class="reference internal" href="../library/logging.html#logging.Filter.filter" title="logging.Filter.filter"><code class="xref py py-meth docutils literal">filter()</code></a> method is called.</li>

1221

</ul>

1222

The first approach would be a little unwieldy in the scenario where (say)

1223

several different libraries wanted to do different things. Each would attempt

1224

to set its own <a class="reference internal" href="../library/logging.html#logging.Logger" title="logging.Logger"><code class="xref py py-class docutils literal">Logger</code></a> subclass, and the one which did this last would

1225

win.

1226

The second approach works reasonably well for many cases, but does not allow

1227

you to e.g. use a specialized subclass of <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a>. Library

1228

developers can set a suitable filter on their loggers, but they would have to

1229

remember to do this every time they introduced a new logger (which they would

1230

do simply by adding new packages or modules and doing

1231

1232

</pre></div>

1233

</div>

1234

at module level). It’s probably one too many things to think about. Developers

1235

could also add the filter to a <a class="reference internal" href="../library/logging.handlers.html#logging.NullHandler" title="logging.NullHandler"><code class="xref py py-class docutils literal">NullHandler</code></a> attached to their

1236

top-level logger, but this would not be invoked if an application developer

1237

attached a handler to a lower-level library logger – so output from that

1238

handler would not reflect the intentions of the library developer.

1239

In Python 3.2 and later, <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a> creation is done through a

1240

factory, which you can specify. The factory is just a callable you can set with

1241

<a class="reference internal" href="../library/logging.html#logging.setLogRecordFactory" title="logging.setLogRecordFactory"><code class="xref py py-func docutils literal">setLogRecordFactory()</code></a>, and interrogate with

1242

<a class="reference internal" href="../library/logging.html#logging.getLogRecordFactory" title="logging.getLogRecordFactory"><code class="xref py py-func docutils literal">getLogRecordFactory()</code></a>. The factory is invoked with the same

1243

signature as the <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a> constructor, as <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a>

1244

is the default setting for the factory.

1245

This approach allows a custom factory to control all aspects of LogRecord

1246

creation. For example, you could return a subclass, or just add some additional

1247

attributes to the record once created, using a pattern similar to this:

1248

<div class="highlight-python3"><div class="highlight"><pre>old_factory = logging.getLogRecordFactory()

1249

1250

def record_factory(*args, **kwargs):

1251

record = old_factory(*args, **kwargs)

1252

record.custom_attribute = 0xdecafbad

1253

return record

1254

1255

logging.setLogRecordFactory(record_factory)

1256

</pre></div>

1257

</div>

1258

This pattern allows different libraries to chain factories together, and as

1259

long as they don’t overwrite each other’s attributes or unintentionally

1260

overwrite the attributes provided as standard, there should be no surprises.

1261

However, it should be borne in mind that each link in the chain adds run-time

1262

overhead to all logging operations, and the technique should only be used when

1263

the use of a <a class="reference internal" href="../library/logging.html#logging.Filter" title="logging.Filter"><code class="xref py py-class docutils literal">Filter</code></a> does not provide the desired result.

1264

</div>

1265

1266

<h2>Subclassing QueueHandler - a ZeroMQ example<a class="headerlink" href="#subclassing-queuehandler-a-zeromq-example" title="Permalink to this headline">¶</a></h2>

1267

You can use a <code class="xref py py-class docutils literal">QueueHandler</code> subclass to send messages to other kinds

1268

of queues, for example a ZeroMQ ‘publish’ socket. In the example below,the

1269

socket is created separately and passed to the handler (as its ‘queue’):

1270

<div class="highlight-python3"><div class="highlight"><pre>import zmq # using pyzmq, the Python binding for ZeroMQ

1271

import json # for serializing records portably

1272

1273

ctx = zmq.Context()

1274

sock = zmq.Socket(ctx, zmq.PUB) # or zmq.PUSH, or other suitable value

1275

sock.bind('tcp://*:5556') # or wherever

1276

1277

class ZeroMQSocketHandler(QueueHandler):

1278

def enqueue(self, record):

1279

data = json.dumps(record.__dict__)

1280

self.queue.send(data)

1281

1282

handler = ZeroMQSocketHandler(sock)

1283

</pre></div>

1284

</div>

1285

Of course there are other ways of organizing this, for example passing in the

1286

data needed by the handler to create the socket:

1287

<div class="highlight-python3"><div class="highlight"><pre>class ZeroMQSocketHandler(QueueHandler):

1288

def __init__(self, uri, socktype=zmq.PUB, ctx=None):

1289

self.ctx = ctx or zmq.Context()

1290

socket = zmq.Socket(self.ctx, socktype)

1291

socket.bind(uri)

1292

QueueHandler.__init__(self, socket)

1293

1294

def enqueue(self, record):

1295

1296

1297

1298

def close(self):

1299

self.queue.close()

1300

</pre></div>

1301

</div>

1302

</div>

1303

1304

<h2>Subclassing QueueListener - a ZeroMQ example<a class="headerlink" href="#subclassing-queuelistener-a-zeromq-example" title="Permalink to this headline">¶</a></h2>

1305

You can also subclass <code class="xref py py-class docutils literal">QueueListener</code> to get messages from other kinds

1306

of queues, for example a ZeroMQ ‘subscribe’ socket. Here’s an example:

1307

<div class="highlight-python3"><div class="highlight"><pre>class ZeroMQSocketListener(QueueListener):

1308

def __init__(self, uri, *handlers, **kwargs):

1309

self.ctx = kwargs.get('ctx') or zmq.Context()

1310

1311

socket.setsockopt(zmq.SUBSCRIBE, '') # subscribe to everything

1312

socket.connect(uri)

1313

1314

def dequeue(self):

1315

msg = self.queue.recv()

1316

return logging.makeLogRecord(json.loads(msg))

1317

</pre></div>

1318

</div>

1319

1320

See also

1321

1322

<dt>Module <a class="reference internal" href="../library/logging.html#module-logging" title="logging: Flexible event logging system for applications."><code class="xref py py-mod docutils literal">logging</code></a></dt>

1323

<dd>API reference for the logging module.</dd>

1324

<dt>Module <a class="reference internal" href="../library/logging.config.html#module-logging.config" title="logging.config: Configuration of the logging module."><code class="xref py py-mod docutils literal">logging.config</code></a></dt>

1325

<dd>Configuration API for the logging module.</dd>

1326

<dt>Module <a class="reference internal" href="../library/logging.handlers.html#module-logging.handlers" title="logging.handlers: Handlers for the logging module."><code class="xref py py-mod docutils literal">logging.handlers</code></a></dt>

1327

<dd>Useful handlers included with the logging module.</dd>

1328

</dl>

1329

<a class="reference internal" href="logging.html#logging-basic-tutorial">A basic logging tutorial</a>

1330

<a class="reference internal" href="logging.html#logging-advanced-tutorial">A more advanced logging tutorial</a>

1331

</div>

1332

</div>

1333

1334

<h2>An example dictionary-based configuration<a class="headerlink" href="#an-example-dictionary-based-configuration" title="Permalink to this headline">¶</a></h2>

1335

Below is an example of a logging configuration dictionary - it’s taken from

1336

the <a class="reference external" href="https://docs.djangoproject.com/en/1.9/topics/logging/#configuring-logging">documentation on the Django project</a>.

1337

This dictionary is passed to <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a> to put the configuration into effect:

1338

<div class="highlight-python3"><div class="highlight"><pre>LOGGING = {

1339

'version': 1,

1340

'disable_existing_loggers': True,

1341

'formatters': {

1342

'verbose': {

1343

'format': '%(levelname)s %(asctime)s %(module)s %(process)d %(thread)d %(message)s'

1344

},

1345

'simple': {

1346

1347

},

1348

},

1349

'filters': {

1350

'special': {

1351

'()': 'project.logging.SpecialFilter',

1352

'foo': 'bar',

1353

}

1354

},

1355

'handlers': {

1356

'null': {

1357

'level':'DEBUG',

1358

'class':'django.utils.log.NullHandler',

1359

},

1360

'console':{

1361

'level':'DEBUG',

1362

'class':'logging.StreamHandler',

1363

'formatter': 'simple'

1364

},

1365

'mail_admins': {

1366

'level': 'ERROR',

1367

'class': 'django.utils.log.AdminEmailHandler',

1368

'filters': ['special']

1369

}

1370

},

1371

'loggers': {

1372

'django': {

1373

'handlers':['null'],

1374

'propagate': True,

1375

'level':'INFO',

1376

},

1377

'django.request': {

1378

'handlers': ['mail_admins'],

1379

'level': 'ERROR',

1380

'propagate': False,

1381

},

1382

'myproject.custom': {

1383

1384

'level': 'INFO',

1385

'filters': ['special']

1386

}

1387

}

1388

}

1389

</pre></div>

1390

</div>

1391

For more information about this configuration, you can see the <a class="reference external" href="https://docs.djangoproject.com/en/1.9/topics/logging/#configuring-logging">relevant

1392

section</a>

1393

of the Django documentation.

1394

</div>

1395

1396

<h2>Using a rotator and namer to customize log rotation processing<a class="headerlink" href="#using-a-rotator-and-namer-to-customize-log-rotation-processing" title="Permalink to this headline">¶</a></h2>

1397

An example of how you can define a namer and rotator is given in the following

1398

snippet, which shows zlib-based compression of the log file:

1399

<div class="highlight-python3"><div class="highlight"><pre>def namer(name):

1400

return name + ".gz"

1401

1402

def rotator(source, dest):

1403

with open(source, "rb") as sf:

1404

data = sf.read()

1405

compressed = zlib.compress(data, 9)

1406

with open(dest, "wb") as df:

1407

df.write(compressed)

1408

os.remove(source)

1409

1410

rh = logging.handlers.RotatingFileHandler(...)

1411

rh.rotator = rotator

1412

rh.namer = namer

1413

</pre></div>

1414

</div>

1415

These are not “true” .gz files, as they are bare compressed data, with no

1416

“container” such as you’d find in an actual gzip file. This snippet is just

1417

for illustration purposes.

1418

</div>

1419

1420

<h2>A more elaborate multiprocessing example<a class="headerlink" href="#a-more-elaborate-multiprocessing-example" title="Permalink to this headline">¶</a></h2>

1421

The following working example shows how logging can be used with multiprocessing

1422

using configuration files. The configurations are fairly simple, but serve to

1423

illustrate how more complex ones could be implemented in a real multiprocessing

1424

scenario.

1425

In the example, the main process spawns a listener process and some worker

1426

processes. Each of the main process, the listener and the workers have three

1427

separate configurations (the workers all share the same configuration). We can

1428

see logging in the main process, how the workers log to a QueueHandler and how

1429

the listener implements a QueueListener and a more complex logging

1430

configuration, and arranges to dispatch events received via the queue to the

1431

handlers specified in the configuration. Note that these configurations are

1432

purely illustrative, but you should be able to adapt this example to your own

1433

scenario.

1434

Here’s the script - the docstrings and the comments hopefully explain how it

1435

works:

1436

<div class="highlight-python3"><div class="highlight"><pre>import logging

1437

import logging.config

1438

import logging.handlers

1439

from multiprocessing import Process, Queue, Event, current_process

1440

import os

1441

import random

1442

import time

1443

1444

class MyHandler:

1445

"""

1446

A simple handler for logging events. It runs in the listener process and

1447

dispatches events to loggers based on the name in the received record,

1448

which then get dispatched, by the logging system, to the handlers

1449

configured for those loggers.

1450

"""

1451

def handle(self, record):

1452

logger = logging.getLogger(record.name)

1453

# The process name is transformed just to show that it's the listener

1454

# doing the logging to files and console

1455

record.processName = '%s (for %s)' % (current_process().name, record.processName)

1456

logger.handle(record)

1457

1458

def listener_process(q, stop_event, config):

1459

"""

1460

This could be done in the main process, but is just done in a separate

1461

process for illustrative purposes.

1462

1463

This initialises logging according to the specified configuration,

1464

starts the listener and waits for the main process to signal completion

1465

via the event. The listener is then stopped, and the process exits.

1466

"""

1467

1468

listener = logging.handlers.QueueListener(q, MyHandler())

1469

listener.start()

1470

if os.name == 'posix':

1471

# On POSIX, the setup logger will have been configured in the

1472

# parent process, but should have been disabled following the

1473

# dictConfig call.

1474

# On Windows, since fork isn't used, the setup logger won't

1475

# exist in the child, so it would be created and the message

1476

# would appear - hence the "if posix" clause.

1477

1478

logger.critical('Should not appear, because of disabled logger ...')

1479

stop_event.wait()

1480

listener.stop()

1481

1482

def worker_process(config):

1483

"""

1484

A number of these are spawned for the purpose of illustration. In

1485

practice, they could be a heterogeneous bunch of processes rather than

1486

ones which are identical to each other.

1487

1488

This initialises logging according to the specified configuration,

1489

and logs a hundred messages with random levels to randomly selected

1490

loggers.

1491

1492

A small sleep is added to allow other processes a chance to run. This

1493

is not strictly needed, but it mixes the output from the different

1494

processes a bit more than if it's left out.

1495

"""

1496

1497

1498

logging.CRITICAL]

1499

1500

'spam', 'spam.ham', 'spam.ham.eggs']

1501

if os.name == 'posix':

1502

# On POSIX, the setup logger will have been configured in the

1503

# parent process, but should have been disabled following the

1504

# dictConfig call.

1505

# On Windows, since fork isn't used, the setup logger won't

1506

# exist in the child, so it would be created and the message

1507

# would appear - hence the "if posix" clause.

1508

1509

1510

for i in range(100):

1511

1512

logger = logging.getLogger(random.choice(loggers))

1513

1514

time.sleep(0.01)

1515

1516

def main():

1517

q = Queue()

1518

# The main process gets a simple configuration which prints to the console.

1519

config_initial = {

1520

'version': 1,

1521

'formatters': {

1522

'detailed': {

1523

'class': 'logging.Formatter',

1524

1525

}

1526

},

1527

'handlers': {

1528

'console': {

1529

'class': 'logging.StreamHandler',

1530

'level': 'INFO',

1531

},

1532

},

1533

'root': {

1534

'level': 'DEBUG',

1535

'handlers': ['console']

1536

},

1537

}

1538

# The worker process configuration is just a QueueHandler attached to the

1539

# root logger, which allows all messages to be sent to the queue.

1540

# We disable existing loggers to disable the "setup" logger used in the

1541

# parent process. This is needed on POSIX because the logger will

1542

# be there in the child following a fork().

1543

config_worker = {

1544

'version': 1,

1545

'disable_existing_loggers': True,

1546

'handlers': {

1547

'queue': {

1548

'class': 'logging.handlers.QueueHandler',

1549

'queue': q,

1550

},

1551

},

1552

'root': {

1553

'level': 'DEBUG',

1554

'handlers': ['queue']

1555

},

1556

}

1557

# The listener process configuration shows that the full flexibility of

1558

# logging configuration is available to dispatch events to handlers however

1559

# you want.

1560

# We disable existing loggers to disable the "setup" logger used in the

1561

# parent process. This is needed on POSIX because the logger will

1562

# be there in the child following a fork().

1563

config_listener = {

1564

'version': 1,

1565

'disable_existing_loggers': True,

1566

'formatters': {

1567

'detailed': {

1568

'class': 'logging.Formatter',

1569

1570

},

1571

'simple': {

1572

'class': 'logging.Formatter',

1573

'format': '%(name)-15s %(levelname)-8s %(processName)-10s %(message)s'

1574

}

1575

},

1576

'handlers': {

1577

'console': {

1578

'class': 'logging.StreamHandler',

1579

'level': 'INFO',

1580

'formatter': 'simple',

1581

},

1582

'file': {

1583

'class': 'logging.FileHandler',

1584

'filename': 'mplog.log',

1585

'mode': 'w',

1586

'formatter': 'detailed',

1587

},

1588

'foofile': {

1589

'class': 'logging.FileHandler',

1590

'filename': 'mplog-foo.log',

1591

'mode': 'w',

1592

'formatter': 'detailed',

1593

},

1594

'errors': {

1595

'class': 'logging.FileHandler',

1596

'filename': 'mplog-errors.log',

1597

'mode': 'w',

1598

'level': 'ERROR',

1599

'formatter': 'detailed',

1600

},

1601

},

1602

'loggers': {

1603

'foo': {

1604

'handlers': ['foofile']

1605

}

1606

},

1607

'root': {

1608

'level': 'DEBUG',

1609

1610

},

1611

}

1612

# Log some initial events, just to show that logging in the parent works

1613

# normally.

1614

1615

1616

logger.info('About to create workers ...')

1617

workers = []

1618

for i in range(5):

1619

1620

args=(config_worker,))

1621

workers.append(wp)

1622

wp.start()

1623

logger.info('Started worker: %s', wp.name)

1624

logger.info('About to create listener ...')

1625

stop_event = Event()

1626

lp = Process(target=listener_process, name='listener',

1627

args=(q, stop_event, config_listener))

1628

lp.start()

1629

logger.info('Started listener')

1630

# We now hang around for the workers to finish their work.

1631

for wp in workers:

1632

wp.join()

1633

# Workers all done, listening can now stop.

1634

# Logging in the parent still works normally.

1635

logger.info('Telling listener to stop ...')

1636

stop_event.set()

1637

lp.join()

1638

logger.info('All done.')

1639

1640

if __name__ == '__main__':

1641

main()

1642

</pre></div>

1643

</div>

1644

</div>

1645

1646

<h2>Inserting a BOM into messages sent to a SysLogHandler<a class="headerlink" href="#inserting-a-bom-into-messages-sent-to-a-sysloghandler" title="Permalink to this headline">¶</a></h2>

1647

<a class="reference external" href="https://tools.ietf.org/html/rfc5424">RFC 5424</a> requires that a

1648

Unicode message be sent to a syslog daemon as a set of bytes which have the

1649

following structure: an optional pure-ASCII component, followed by a UTF-8 Byte

1650

Order Mark (BOM), followed by Unicode encoded using UTF-8. (See the <a class="reference external" href="https://tools.ietf.org/html/rfc5424#section-6">relevant

1651

section of the specification</a>.)

1652

In Python 3.1, code was added to

1653

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.SysLogHandler" title="logging.handlers.SysLogHandler"><code class="xref py py-class docutils literal">SysLogHandler</code></a> to insert a BOM into the message, but

1654

unfortunately, it was implemented incorrectly, with the BOM appearing at the

1655

beginning of the message and hence not allowing any pure-ASCII component to

1656

appear before it.

1657

As this behaviour is broken, the incorrect BOM insertion code is being removed

1658

from Python 3.2.4 and later. However, it is not being replaced, and if you

1659

want to produce RFC 5424-compliant messages which include a BOM, an optional

1660

pure-ASCII sequence before it and arbitrary Unicode after it, encoded using

1661

UTF-8, then you need to do the following:

1662

1663

<li>Attach a <a class="reference internal" href="../library/logging.html#logging.Formatter" title="logging.Formatter"><code class="xref py py-class docutils literal">Formatter</code></a> instance to your

1664

1665

such as:

1666

<div class="highlight-python3"><div class="highlight"><pre>'ASCII section\ufeffUnicode section'

1667

</pre></div>

1668

</div>

1669

The Unicode code point U+FEFF, when encoded using UTF-8, will be

1670

encoded as a UTF-8 BOM – the byte-string <code class="docutils literal">b'\xef\xbb\xbf'</code>.

1671

</li>

1672

<li>Replace the ASCII section with whatever placeholders you like, but make sure

1673

that the data that appears in there after substitution is always ASCII (that

1674

way, it will remain unchanged after UTF-8 encoding).

1675

</li>

1676

<li>Replace the Unicode section with whatever placeholders you like; if the data

1677

which appears there after substitution contains characters outside the ASCII

1678

range, that’s fine – it will be encoded using UTF-8.

1679

</li>

1680

</ol>

1681

The formatted message will be encoded using UTF-8 encoding by

1682

<code class="docutils literal">SysLogHandler</code>. If you follow the above rules, you should be able to produce

1683

RFC 5424-compliant messages. If you don’t, logging may not complain, but your

1684

messages will not be RFC 5424-compliant, and your syslog daemon may complain.

1685

</div>

1686

1687

<h2>Implementing structured logging<a class="headerlink" href="#implementing-structured-logging" title="Permalink to this headline">¶</a></h2>

1688

Although most logging messages are intended for reading by humans, and thus not

1689

readily machine-parseable, there might be cirumstances where you want to output

1690

messages in a structured format which is capable of being parsed by a program

1691

(without needing complex regular expressions to parse the log message). This is

1692

straightforward to achieve using the logging package. There are a number of

1693

ways in which this could be achieved, but the following is a simple approach

1694

which uses JSON to serialise the event in a machine-parseable manner:

1695

<div class="highlight-python3"><div class="highlight"><pre>import json

1696

import logging

1697

1698

class StructuredMessage(object):

1699

def __init__(self, message, **kwargs):

1700

self.message = message

1701

self.kwargs = kwargs

1702

1703

def __str__(self):

1704

return '%s >>> %s' % (self.message, json.dumps(self.kwargs))

1705

1706

_ = StructuredMessage # optional, to improve readability

1707

1708

logging.basicConfig(level=logging.INFO, format='%(message)s')

1709

logging.info(_('message 1', foo='bar', bar='baz', num=123, fnum=123.456))

1710

</pre></div>

1711

</div>

1712

If the above script is run, it prints:

1713

<div class="highlight-python3"><div class="highlight"><pre>message 1 >>> {"fnum": 123.456, "num": 123, "bar": "baz", "foo": "bar"}

1714

</pre></div>

1715

</div>

1716

Note that the order of items might be different according to the version of

1717

Python used.

1718

If you need more specialised processing, you can use a custom JSON encoder,

1719

as in the following complete example:

1720

<div class="highlight-python3"><div class="highlight"><pre>from __future__ import unicode_literals

1721

1722

import json

1723

import logging

1724

1725

# This next bit is to ensure the script runs unchanged on 2.x and 3.x

1726

try:

1727

unicode

1728

except NameError:

1729

unicode = str

1730

1731

class Encoder(json.JSONEncoder):

1732

def default(self, o):

1733

if isinstance(o, set):

1734

return tuple(o)

1735

elif isinstance(o, unicode):

1736

return o.encode('unicode_escape').decode('ascii')

1737

return super(Encoder, self).default(o)

1738

1739

class StructuredMessage(object):

1740

1741

self.message = message

1742

self.kwargs = kwargs

1743

1744

def __str__(self):

1745

s = Encoder().encode(self.kwargs)

1746

1747

1748

_ = StructuredMessage # optional, to improve readability

1749

1750

def main():

1751

logging.basicConfig(level=logging.INFO, format='%(message)s')

1752

logging.info(_('message 1', set_value={1, 2, 3}, snowman='\u2603'))

1753

1754

if __name__ == '__main__':

1755

main()

1756

</pre></div>

1757

</div>

1758

When the above script is run, it prints:

1759

<div class="highlight-python3"><div class="highlight"><pre>message 1 >>> {"snowman": "\u2603", "set_value": [1, 2, 3]}

1760

</pre></div>

1761

</div>

1762

Note that the order of items might be different according to the version of

1763

Python used.

1764

</div>

1765

1766

<h2>Customizing handlers with <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a><a class="headerlink" href="#customizing-handlers-with-dictconfig" title="Permalink to this headline">¶</a></h2>

1767

There are times when you want to customize logging handlers in particular ways,

1768

and if you use <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a> you may be able to do this without

1769

subclassing. As an example, consider that you may want to set the ownership of a

1770

log file. On POSIX, this is easily done using <a class="reference internal" href="../library/shutil.html#shutil.chown" title="shutil.chown"><code class="xref py py-func docutils literal">shutil.chown()</code></a>, but the file

1771

handlers in the stdlib don’t offer built-in support. You can customize handler

1772

creation using a plain function such as:

1773

<div class="highlight-python3"><div class="highlight"><pre>def owned_file_handler(filename, mode='a', encoding=None, owner=None):

1774

if owner:

1775

if not os.path.exists(filename):

1776

open(filename, 'a').close()

1777

shutil.chown(filename, *owner)

1778

return logging.FileHandler(filename, mode, encoding)

1779

</pre></div>

1780

</div>

1781

You can then specify, in a logging configuration passed to <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a>,

1782

that a logging handler be created by calling this function:

1783

<div class="highlight-python3"><div class="highlight"><pre>LOGGING = {

1784

'version': 1,

1785

'disable_existing_loggers': False,

1786

'formatters': {

1787

'default': {

1788

'format': '%(asctime)s %(levelname)s %(name)s %(message)s'

1789

},

1790

},

1791

'handlers': {

1792

'file':{

1793

# The values below are popped from this dictionary and

1794

# used to create the handler, set the handler's level and

1795

# its formatter.

1796

'()': owned_file_handler,

1797

'level':'DEBUG',

1798

'formatter': 'default',

1799

# The values below are passed to the handler creator callable

1800

# as keyword arguments.

1801

'owner': ['pulse', 'pulse'],

1802

'filename': 'chowntest.log',

1803

'mode': 'w',

1804

'encoding': 'utf-8',

1805

},

1806

},

1807

'root': {

1808

'handlers': ['file'],

1809

'level': 'DEBUG',

1810

},

1811

}

1812

</pre></div>

1813

</div>

1814

In this example I am setting the ownership using the <code class="docutils literal">pulse</code> user and group,

1815

just for the purposes of illustration. Putting it together into a working

1816

script, <code class="docutils literal">chowntest.py</code>:

1817

<div class="highlight-python3"><div class="highlight"><pre>import logging, logging.config, os, shutil

1818

1819

def owned_file_handler(filename, mode='a', encoding=None, owner=None):

1820

if owner:

1821

1822

1823

1824

1825

1826

LOGGING = {

1827

'version': 1,

1828

'disable_existing_loggers': False,

1829

'formatters': {

1830

'default': {

1831

'format': '%(asctime)s %(levelname)s %(name)s %(message)s'

1832

},

1833

},

1834

'handlers': {

1835

'file':{

1836

# The values below are popped from this dictionary and

1837

# used to create the handler, set the handler's level and

1838

# its formatter.

1839

'()': owned_file_handler,

1840

'level':'DEBUG',

1841

'formatter': 'default',

1842

# The values below are passed to the handler creator callable

1843

# as keyword arguments.

1844

'owner': ['pulse', 'pulse'],

1845

'filename': 'chowntest.log',

1846

'mode': 'w',

1847

'encoding': 'utf-8',

1848

},

1849

},

1850

'root': {

1851

'handlers': ['file'],

1852

'level': 'DEBUG',

1853

},

1854

}

1855

1856

1857

1858

logger.debug('A debug message')

1859

</pre></div>

1860

</div>

1861

To run this, you will probably need to run as <code class="docutils literal">root</code>:

1862

<div class="highlight-shell-session"><div class="highlight"><pre>$ sudo python3.3 chowntest.py

1863

$ cat chowntest.log

1864

2013-11-05 09:34:51,128 DEBUG mylogger A debug message

1865

$ ls -l chowntest.log

1866

-rw-r--r-- 1 pulse pulse 55 2013-11-05 09:34 chowntest.log

1867

</pre></div>

1868

</div>

1869

Note that this example uses Python 3.3 because that’s where <a class="reference internal" href="../library/shutil.html#shutil.chown" title="shutil.chown"><code class="xref py py-func docutils literal">shutil.chown()</code></a>

1870

makes an appearance. This approach should work with any Python version that

1871

supports <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a> - namely, Python 2.7, 3.2 or later. With pre-3.3

1872

versions, you would need to implement the actual ownership change using e.g.

1873

<a class="reference internal" href="../library/os.html#os.chown" title="os.chown"><code class="xref py py-func docutils literal">os.chown()</code></a>.

1874

In practice, the handler-creating function may be in a utility module somewhere

1875

in your project. Instead of the line in the configuration:

1876

<div class="highlight-python3"><div class="highlight"><pre>'()': owned_file_handler,

1877

</pre></div>

1878

</div>

1879

you could use e.g.:

1880

<div class="highlight-python3"><div class="highlight"><pre>'()': 'ext://project.util.owned_file_handler',

1881

</pre></div>

1882

</div>

1883

where <code class="docutils literal">project.util</code> can be replaced with the actual name of the package

1884

where the function resides. In the above working script, using

1885

<code class="docutils literal">'ext://__main__.owned_file_handler'</code> should work. Here, the actual callable

1886

is resolved by <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a> from the <code class="docutils literal">ext://</code> specification.

1887

This example hopefully also points the way to how you could implement other

1888

types of file change - e.g. setting specific POSIX permission bits - in the

1889

same way, using <a class="reference internal" href="../library/os.html#os.chmod" title="os.chmod"><code class="xref py py-func docutils literal">os.chmod()</code></a>.

1890

Of course, the approach could also be extended to types of handler other than a

1891

<a class="reference internal" href="../library/logging.handlers.html#logging.FileHandler" title="logging.FileHandler"><code class="xref py py-class docutils literal">FileHandler</code></a> - for example, one of the rotating file handlers,

1892

or a different type of handler altogether.

1893

</div>

1894

1895

<h2>Using particular formatting styles throughout your application<a class="headerlink" href="#using-particular-formatting-styles-throughout-your-application" title="Permalink to this headline">¶</a></h2>

1896

In Python 3.2, the <a class="reference internal" href="../library/logging.html#logging.Formatter" title="logging.Formatter"><code class="xref py py-class docutils literal">Formatter</code></a> gained a <code class="docutils literal">style</code> keyword

1897

parameter which, while defaulting to <code class="docutils literal">%</code> for backward compatibility, allowed

1898

the specification of <code class="docutils literal">{</code> or <code class="docutils literal">$</code> to support the formatting approaches

1899

supported by <a class="reference internal" href="../library/stdtypes.html#str.format" title="str.format"><code class="xref py py-meth docutils literal">str.format()</code></a> and <a class="reference internal" href="../library/string.html#string.Template" title="string.Template"><code class="xref py py-class docutils literal">string.Template</code></a>. Note that this

1900

governs the formatting of logging messages for final output to logs, and is

1901

completely orthogonal to how an individual logging message is constructed.

1902

Logging calls (<a class="reference internal" href="../library/logging.html#logging.Logger.debug" title="logging.Logger.debug"><code class="xref py py-meth docutils literal">debug()</code></a>, <a class="reference internal" href="../library/logging.html#logging.Logger.info" title="logging.Logger.info"><code class="xref py py-meth docutils literal">info()</code></a> etc.) only take

1903

positional parameters for the actual logging message itself, with keyword

1904

parameters used only for determining options for how to handle the logging call

1905

(e.g. the <code class="docutils literal">exc_info</code> keyword parameter to indicate that traceback information

1906

should be logged, or the <code class="docutils literal">extra</code> keyword parameter to indicate additional

1907

contextual information to be added to the log). So you cannot directly make

1908

logging calls using <a class="reference internal" href="../library/stdtypes.html#str.format" title="str.format"><code class="xref py py-meth docutils literal">str.format()</code></a> or <a class="reference internal" href="../library/string.html#string.Template" title="string.Template"><code class="xref py py-class docutils literal">string.Template</code></a> syntax,

1909

because internally the logging package uses %-formatting to merge the format

1910

string and the variable arguments. There would no changing this while preserving

1911

backward compatibility, since all logging calls which are out there in existing

1912

code will be using %-format strings.

1913

There have been suggestions to associate format styles with specific loggers,

1914

but that approach also runs into backward compatibility problems because any

1915

existing code could be using a given logger name and using %-formatting.

1916

For logging to work interoperably between any third-party libraries and your

1917

code, decisions about formatting need to be made at the level of the

1918

individual logging call. This opens up a couple of ways in which alternative

1919

formatting styles can be accommodated.

1920

1921

<h3>Using LogRecord factories<a class="headerlink" href="#using-logrecord-factories" title="Permalink to this headline">¶</a></h3>

1922

In Python 3.2, along with the <a class="reference internal" href="../library/logging.html#logging.Formatter" title="logging.Formatter"><code class="xref py py-class docutils literal">Formatter</code></a> changes mentioned

1923

above, the logging package gained the ability to allow users to set their own

1924

<a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a> subclasses, using the <a class="reference internal" href="../library/logging.html#logging.setLogRecordFactory" title="logging.setLogRecordFactory"><code class="xref py py-func docutils literal">setLogRecordFactory()</code></a> function.

1925

You can use this to set your own subclass of <a class="reference internal" href="../library/logging.html#logging.LogRecord" title="logging.LogRecord"><code class="xref py py-class docutils literal">LogRecord</code></a>, which does the

1926

Right Thing by overriding the <a class="reference internal" href="../library/logging.html#logging.LogRecord.getMessage" title="logging.LogRecord.getMessage"><code class="xref py py-meth docutils literal">getMessage()</code></a> method. The base

1927

class implementation of this method is where the <code class="docutils literal">msg % args</code> formatting

1928

happens, and where you can substitute your alternate formatting; however, you

1929

should be careful to support all formatting styles and allow %-formatting as

1930

the default, to ensure interoperability with other code. Care should also be

1931

taken to call <code class="docutils literal">str(self.msg)</code>, just as the base implementation does.

1932

Refer to the reference documentation on <a class="reference internal" href="../library/logging.html#logging.setLogRecordFactory" title="logging.setLogRecordFactory"><code class="xref py py-func docutils literal">setLogRecordFactory()</code></a> and

1933

1934

</div>

1935

1936

<h3>Using custom message objects<a class="headerlink" href="#using-custom-message-objects" title="Permalink to this headline">¶</a></h3>

1937

There is another, perhaps simpler way that you can use {}- and $- formatting to

1938

construct your individual log messages. You may recall (from

1939

<a class="reference internal" href="logging.html#arbitrary-object-messages">Using arbitrary objects as messages</a>) that when logging you can use an arbitrary

1940

object as a message format string, and that the logging package will call

1941

<a class="reference internal" href="../library/stdtypes.html#str" title="str"><code class="xref py py-func docutils literal">str()</code></a> on that object to get the actual format string. Consider the

1942

following two classes:

1943

<div class="highlight-python3"><div class="highlight"><pre>class BraceMessage(object):

1944

1945

self.fmt = fmt

1946

self.args = args

1947

self.kwargs = kwargs

1948

1949

def __str__(self):

1950

1951

1952

class DollarMessage(object):

1953

def __init__(self, fmt, **kwargs):

1954

self.fmt = fmt

1955

self.kwargs = kwargs

1956

1957

def __str__(self):

1958

from string import Template

1959

1960

</pre></div>

1961

</div>

1962

Either of these can be used in place of a format string, to allow {}- or

1963

$-formatting to be used to build the actual “message” part which appears in the

1964

formatted log output in place of “%(message)s” or “{message}” or “$message”.

1965

If you find it a little unwieldy to use the class names whenever you want to log

1966

something, you can make it more palatable if you use an alias such as <code class="docutils literal">M</code> or

1967

<code class="docutils literal">_</code> for the message (or perhaps <code class="docutils literal">__</code>, if you are using <code class="docutils literal">_</code> for

1968

localization).

1969

Examples of this approach are given below. Firstly, formatting with

1970

<a class="reference internal" href="../library/stdtypes.html#str.format" title="str.format"><code class="xref py py-meth docutils literal">str.format()</code></a>:

1971

<div class="highlight-python3"><div class="highlight"><pre>>>> __ = BraceMessage

1972

>>> print(__('Message with {0} {1}', 2, 'placeholders'))

1973

Message with 2 placeholders

1974

>>> class Point: pass

1975

...

1976

>>> p = Point()

1977

>>> p.x = 0.5

1978

>>> p.y = 0.5

1979

>>> print(__('Message with coordinates: ({point.x:.2f}, {point.y:.2f})', point=p))

1980

Message with coordinates: (0.50, 0.50)

1981

</pre></div>

1982

</div>

1983

Secondly, formatting with <a class="reference internal" href="../library/string.html#string.Template" title="string.Template"><code class="xref py py-class docutils literal">string.Template</code></a>:

1984

<div class="highlight-python3"><div class="highlight"><pre>>>> __ = DollarMessage

1985

>>> print(__('Message with $num $what', num=2, what='placeholders'))

1986

Message with 2 placeholders

1987

>>>

1988

</pre></div>

1989

</div>

1990

One thing to note is that you pay no significant performance penalty with this

1991

approach: the actual formatting happens not when you make the logging call, but

1992

when (and if) the logged message is actually about to be output to a log by a

1993

handler. So the only slightly unusual thing which might trip you up is that the

1994

parentheses go around the format string and the arguments, not just the format

1995

string. That’s because the __ notation is just syntax sugar for a constructor

1996

call to one of the <code class="docutils literal">XXXMessage</code> classes shown above.

1997

</div>

1998

</div>

1999

2000

<h2>Configuring filters with <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a><a class="headerlink" href="#configuring-filters-with-dictconfig" title="Permalink to this headline">¶</a></h2>

2001

You can configure filters using <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a>, though it

2002

might not be obvious at first glance how to do it (hence this recipe). Since

2003

<a class="reference internal" href="../library/logging.html#logging.Filter" title="logging.Filter"><code class="xref py py-class docutils literal">Filter</code></a> is the only filter class included in the standard

2004

library, and it is unlikely to cater to many requirements (it’s only there as a

2005

base class), you will typically need to define your own <a class="reference internal" href="../library/logging.html#logging.Filter" title="logging.Filter"><code class="xref py py-class docutils literal">Filter</code></a>

2006

subclass with an overridden <a class="reference internal" href="../library/logging.html#logging.Filter.filter" title="logging.Filter.filter"><code class="xref py py-meth docutils literal">filter()</code></a> method. To do this,

2007

specify the <code class="docutils literal">()</code> key in the configuration dictionary for the filter,

2008

specifying a callable which will be used to create the filter (a class is the

2009

most obvious, but you can provide any callable which returns a

2010

2011

<div class="highlight-python3"><div class="highlight"><pre>import logging

2012

import logging.config

2013

import sys

2014

2015

class MyFilter(logging.Filter):

2016

def __init__(self, param=None):

2017

self.param = param

2018

2019

def filter(self, record):

2020

if self.param is None:

2021

allow = True

2022

else:

2023

allow = self.param not in record.msg

2024

if allow:

2025

record.msg = 'changed: ' + record.msg

2026

return allow

2027

2028

LOGGING = {

2029

'version': 1,

2030

'filters': {

2031

'myfilter': {

2032

'()': MyFilter,

2033

'param': 'noshow',

2034

}

2035

},

2036

'handlers': {

2037

'console': {

2038

'class': 'logging.StreamHandler',

2039

'filters': ['myfilter']

2040

}

2041

},

2042

'root': {

2043

'level': 'DEBUG',

2044

'handlers': ['console']

2045

},

2046

}

2047

2048

if __name__ == '__main__':

2049

2050

logging.debug('hello')

2051

logging.debug('hello - noshow')

2052

</pre></div>

2053

</div>

2054

This example shows how you can pass configuration data to the callable which

2055

constructs the instance, in the form of keyword parameters. When run, the above

2056

script will print:

2057

<div class="highlight-python3"><div class="highlight"><pre>changed: hello

2058

</pre></div>

2059

</div>

2060

which shows that the filter is working as configured.

2061

A couple of extra points to note:

2062

2063

<li>If you can’t refer to the callable directly in the configuration (e.g. if it

2064

lives in a different module, and you can’t import it directly where the

2065

configuration dictionary is), you can use the form <code class="docutils literal">ext://...</code> as described

2066

in <a class="reference internal" href="../library/logging.config.html#logging-config-dict-externalobj">Access to external objects</a>. For example, you could have used

2067

the text <code class="docutils literal">'ext://__main__.MyFilter'</code> instead of <code class="docutils literal">MyFilter</code> in the above

2068

example.</li>

2069

<li>As well as for filters, this technique can also be used to configure custom

2070

handlers and formatters. See <a class="reference internal" href="../library/logging.config.html#logging-config-dict-userdef">User-defined objects</a> for more

2071

information on how logging supports using user-defined objects in its

2072

configuration, and see the other cookbook recipe <a class="reference internal" href="#custom-handlers">Customizing handlers with dictConfig()</a> above.</li>

2073

</ul>

2074

</div>

2075

2076

<h2>Customized exception formatting<a class="headerlink" href="#customized-exception-formatting" title="Permalink to this headline">¶</a></h2>

2077

There might be times when you want to do customized exception formatting - for

2078

argument’s sake, let’s say you want exactly one line per logged event, even

2079

when exception information is present. You can do this with a custom formatter

2080

class, as shown in the following example:

2081

<div class="highlight-python3"><div class="highlight"><pre>import logging

2082

2083

class OneLineExceptionFormatter(logging.Formatter):

2084

def formatException(self, exc_info):

2085

"""

2086

Format an exception so that it prints on a single line.

2087

"""

2088

result = super(OneLineExceptionFormatter, self).formatException(exc_info)

2089

return repr(result) # or format into one line however you want to

2090

2091

def format(self, record):

2092

s = super(OneLineExceptionFormatter, self).format(record)

2093

if record.exc_text:

2094

s = s.replace('\n', '') + '|'

2095

return s

2096

2097

def configure_logging():

2098

2099

f = OneLineExceptionFormatter('%(asctime)s|%(levelname)s|%(message)s|',

2100

'%d/%m/%Y %H:%M:%S')

2101

fh.setFormatter(f)

2102

root = logging.getLogger()

2103

2104

root.addHandler(fh)

2105

2106

def main():

2107

configure_logging()

2108

logging.info('Sample message')

2109

try:

2110

x = 1 / 0

2111

except ZeroDivisionError as e:

2112

logging.exception('ZeroDivisionError: %s', e)

2113

2114

if __name__ == '__main__':

2115

main()

2116

</pre></div>

2117

</div>

2118

When run, this produces a file with exactly two lines:

2119

<div class="highlight-python3"><div class="highlight"><pre>28/01/2015 07:21:23|INFO|Sample message|

2120

28/01/2015 07:21:23|ERROR|ZeroDivisionError: integer division or modulo by zero|'Traceback (most recent call last):\n File "logtest7.py", line 30, in main\n x = 1 / 0\nZeroDivisionError: integer division or modulo by zero'|

2121

</pre></div>

2122

</div>

2123

While the above treatment is simplistic, it points the way to how exception

2124

information can be formatted to your liking. The <a class="reference internal" href="../library/traceback.html#module-traceback" title="traceback: Print or retrieve a stack traceback."><code class="xref py py-mod docutils literal">traceback</code></a> module may be

2125

helpful for more specialized needs.

2126

</div>

2127

2128

<h2>Speaking logging messages<a class="headerlink" href="#speaking-logging-messages" title="Permalink to this headline">¶</a></h2>

2129

There might be situations when it is desirable to have logging messages rendered

2130

in an audible rather than a visible format. This is easy to do if you have text-

2131

to-speech (TTS) functionality available in your system, even if it doesn’t have

2132

a Python binding. Most TTS systems have a command line program you can run, and

2133

this can be invoked from a handler using <a class="reference internal" href="../library/subprocess.html#module-subprocess" title="subprocess: Subprocess management."><code class="xref py py-mod docutils literal">subprocess</code></a>. It’s assumed here

2134

that TTS command line programs won’t expect to interact with users or take a

2135

long time to complete, and that the frequency of logged messages will be not so

2136

high as to swamp the user with messages, and that it’s acceptable to have the

2137

messages spoken one at a time rather than concurrently, The example implementation

2138

below waits for one message to be spoken before the next is processed, and this

2139

might cause other handlers to be kept waiting. Here is a short example showing

2140

the approach, which assumes that the <code class="docutils literal">espeak</code> TTS package is available:

2141

<div class="highlight-python3"><div class="highlight"><pre>import logging

2142

import subprocess

2143

import sys

2144

2145

class TTSHandler(logging.Handler):

2146

def emit(self, record):

2147

msg = self.format(record)

2148

# Speak slowly in a female English voice

2149

cmd = ['espeak', '-s150', '-ven+f3', msg]

2150

p = subprocess.Popen(cmd, stdout=subprocess.PIPE,

2151

stderr=subprocess.STDOUT)

2152

# wait for the program to finish

2153

p.communicate()

2154

2155

def configure_logging():

2156

h = TTSHandler()

2157

root = logging.getLogger()

2158

root.addHandler(h)

2159

# the default formatter just returns the message

2160

2161

2162

def main():

2163

logging.info('Hello')

2164

logging.debug('Goodbye')

2165

2166

if __name__ == '__main__':

2167

configure_logging()

2168

sys.exit(main())

2169

</pre></div>

2170

</div>

2171

When run, this script should say “Hello” and then “Goodbye” in a female voice.

2172

The above approach can, of course, be adapted to other TTS systems and even

2173

other systems altogether which can process messages via external programs run

2174

from a command line.

2175

</div>

2176

2177

<h2>Buffering logging messages and outputting them conditionally<a class="headerlink" href="#buffering-logging-messages-and-outputting-them-conditionally" title="Permalink to this headline">¶</a></h2>

2178

There might be situations where you want to log messages in a temporary area

2179

and only output them if a certain condition occurs. For example, you may want to

2180

start logging debug events in a function, and if the function completes without

2181

errors, you don’t want to clutter the log with the collected debug information,

2182

but if there is an error, you want all the debug information to be output as well

2183

as the error.

2184

Here is an example which shows how you could do this using a decorator for your

2185

functions where you want logging to behave this way. It makes use of the

2186

<a class="reference internal" href="../library/logging.handlers.html#logging.handlers.MemoryHandler" title="logging.handlers.MemoryHandler"><code class="xref py py-class docutils literal">logging.handlers.MemoryHandler</code></a>, which allows buffering of logged events

2187

until some condition occurs, at which point the buffered events are <code class="docutils literal">flushed</code>

2188

- passed to another handler (the <code class="docutils literal">target</code> handler) for processing. By default,

2189

the <code class="docutils literal">MemoryHandler</code> flushed when its buffer gets filled up or an event whose

2190

level is greater than or equal to a specified threshold is seen. You can use this

2191

recipe with a more specialised subclass of <code class="docutils literal">MemoryHandler</code> if you want custom

2192

flushing behavior.

2193

The example script has a simple function, <code class="docutils literal">foo</code>, which just cycles through

2194

all the logging levels, writing to <code class="docutils literal">sys.stderr</code> to say what level it’s about

2195

to log at, and then actually logging a message at that level. You can pass a

2196

parameter to <code class="docutils literal">foo</code> which, if true, will log at ERROR and CRITICAL levels -

2197

otherwise, it only logs at DEBUG, INFO and WARNING levels.

2198

The script just arranges to decorate <code class="docutils literal">foo</code> with a decorator which will do the

2199

conditional logging that’s required. The decorator takes a logger as a parameter

2200

and attaches a memory handler for the duration of the call to the decorated

2201

function. The decorator can be additionally parameterised using a target handler,

2202

a level at which flushing should occur, and a capacity for the buffer. These

2203

default to a <a class="reference internal" href="../library/logging.handlers.html#logging.StreamHandler" title="logging.StreamHandler"><code class="xref py py-class docutils literal">StreamHandler</code></a> which writes to <code class="docutils literal">sys.stderr</code>,

2204

<code class="docutils literal">logging.ERROR</code> and <code class="docutils literal">100</code> respectively.

2205

Here’s the script:

2206

<div class="highlight-python3"><div class="highlight"><pre>import logging

2207

from logging.handlers import MemoryHandler

2208

import sys

2209

2210

2211

logger.addHandler(logging.NullHandler())

2212

2213

def log_if_errors(logger, target_handler=None, flush_level=None, capacity=None):

2214

if target_handler is None:

2215

target_handler = logging.StreamHandler()

2216

if flush_level is None:

2217

flush_level = logging.ERROR

2218

if capacity is None:

2219

capacity = 100

2220

handler = MemoryHandler(capacity, flushLevel=flush_level, target=target_handler)

2221

2222

def decorator(fn):

2223

def wrapper(*args, **kwargs):

2224

logger.addHandler(handler)

2225

try:

2226

return fn(*args, **kwargs)

2227

except Exception:

2228

logger.exception('call failed')

2229

raise

2230

finally:

2231

super(MemoryHandler, handler).flush()

2232

logger.removeHandler(handler)

2233

return wrapper

2234

2235

return decorator

2236

2237

def write_line(s):

2238

sys.stderr.write('%s\n' % s)

2239

2240

def foo(fail=False):

2241

write_line('about to log at DEBUG ...')

2242

logger.debug('Actually logged at DEBUG')

2243

write_line('about to log at INFO ...')

2244

logger.info('Actually logged at INFO')

2245

write_line('about to log at WARNING ...')

2246

logger.warning('Actually logged at WARNING')

2247

if fail:

2248

write_line('about to log at ERROR ...')

2249

logger.error('Actually logged at ERROR')

2250

write_line('about to log at CRITICAL ...')

2251

logger.critical('Actually logged at CRITICAL')

2252

return fail

2253

2254

decorated_foo = log_if_errors(logger)(foo)

2255

2256

if __name__ == '__main__':

2257

2258

write_line('Calling undecorated foo with False')

2259

assert not foo(False)

2260

write_line('Calling undecorated foo with True')

2261

assert foo(True)

2262

write_line('Calling decorated foo with False')

2263

assert not decorated_foo(False)

2264

write_line('Calling decorated foo with True')

2265

assert decorated_foo(True)

2266

</pre></div>

2267

</div>

2268

When this script is run, the following output should be observed:

2269

<div class="highlight-python3"><div class="highlight"><pre>Calling undecorated foo with False

2270

about to log at DEBUG ...

2271

about to log at INFO ...

2272

about to log at WARNING ...

2273

Calling undecorated foo with True

2274

about to log at DEBUG ...

2275

about to log at INFO ...

2276

about to log at WARNING ...

2277

about to log at ERROR ...

2278

about to log at CRITICAL ...

2279

Calling decorated foo with False

2280

about to log at DEBUG ...

2281

about to log at INFO ...

2282

about to log at WARNING ...

2283

Calling decorated foo with True

2284

about to log at DEBUG ...

2285

about to log at INFO ...

2286

about to log at WARNING ...

2287

about to log at ERROR ...

2288

Actually logged at DEBUG

2289

Actually logged at INFO

2290

Actually logged at WARNING

2291

Actually logged at ERROR

2292

about to log at CRITICAL ...

2293

Actually logged at CRITICAL

2294

</pre></div>

2295

</div>

2296

As you can see, actual logging output only occurs when an event is logged whose

2297

severity is ERROR or greater, but in that case, any previous events at lower

2298

severities are also logged.

2299

You can of course use the conventional means of decoration:

2300

<div class="highlight-python3"><div class="highlight"><pre>@log_if_errors(logger)

2301

def foo(fail=False):

2302

...

2303

</pre></div>

2304

</div>

2305

</div>

2306

2307

<h2>Formatting times using UTC (GMT) via configuration<a class="headerlink" href="#formatting-times-using-utc-gmt-via-configuration" title="Permalink to this headline">¶</a></h2>

2308

Sometimes you want to format times using UTC, which can be done using a class

2309

such as <cite>UTCFormatter</cite>, shown below:

2310

<div class="highlight-python3"><div class="highlight"><pre>import logging

2311

import time

2312

2313

class UTCFormatter(logging.Formatter):

2314

converter = time.gmtime

2315

</pre></div>

2316

</div>

2317

and you can then use the <code class="docutils literal">UTCFormatter</code> in your code instead of

2318

<a class="reference internal" href="../library/logging.html#logging.Formatter" title="logging.Formatter"><code class="xref py py-class docutils literal">Formatter</code></a>. If you want to do that via configuration, you can

2319

use the <a class="reference internal" href="../library/logging.config.html#logging.config.dictConfig" title="logging.config.dictConfig"><code class="xref py py-func docutils literal">dictConfig()</code></a> API with an approach illustrated by

2320

the following complete example:

2321

<div class="highlight-python3"><div class="highlight"><pre>import logging

2322

import logging.config

2323

import time

2324

2325

2326

converter = time.gmtime

2327

2328

LOGGING = {

2329

'version': 1,

2330

'disable_existing_loggers': False,

2331

'formatters': {

2332

'utc': {

2333

'()': UTCFormatter,

2334

2335

},

2336

'local': {

2337

2338

}

2339

},

2340

'handlers': {

2341

'console1': {

2342

'class': 'logging.StreamHandler',

2343

'formatter': 'utc',

2344

},

2345

'console2': {

2346

'class': 'logging.StreamHandler',

2347

'formatter': 'local',

2348

},

2349

},

2350

'root': {

2351

'handlers': ['console1', 'console2'],

2352

}

2353

}

2354

2355

if __name__ == '__main__':

2356

2357

logging.warning('The local time is %s', time.asctime())

2358

</pre></div>

2359

</div>

2360

When this script is run, it should print something like:

2361

<div class="highlight-python3"><div class="highlight"><pre>2015-10-17 12:53:29,501 The local time is Sat Oct 17 13:53:29 2015

2362

2015-10-17 13:53:29,501 The local time is Sat Oct 17 13:53:29 2015

2363

</pre></div>

2364

</div>

2365

showing how the time is formatted both as local time and UTC, one for each

2366

handler.

2367

</div>

2368

2369

<h2>Using a context manager for selective logging<a class="headerlink" href="#using-a-context-manager-for-selective-logging" title="Permalink to this headline">¶</a></h2>

2370

There are times when it would be useful to temporarily change the logging

2371

configuration and revert it back after doing something. For this, a context

2372

manager is the most obvious way of saving and restoring the logging context.

2373

Here is a simple example of such a context manager, which allows you to

2374

optionally change the logging level and add a logging handler purely in the

2375

scope of the context manager:

2376

<div class="highlight-python3"><div class="highlight"><pre>import logging

2377

import sys

2378

2379

class LoggingContext(object):

2380

def __init__(self, logger, level=None, handler=None, close=True):

2381

self.logger = logger

2382

self.level = level

2383

self.handler = handler

2384

self.close = close

2385

2386

def __enter__(self):

2387

if self.level is not None:

2388

self.old_level = self.logger.level

2389

self.logger.setLevel(self.level)

2390

if self.handler:

2391

self.logger.addHandler(self.handler)

2392

2393

def __exit__(self, et, ev, tb):

2394

2395

2396

if self.handler:

2397

self.logger.removeHandler(self.handler)

2398

if self.handler and self.close:

2399

self.handler.close()

2400

# implicit return of None => don't swallow exceptions

2401

</pre></div>

2402

</div>

2403

If you specify a level value, the logger’s level is set to that value in the

2404

scope of the with block covered by the context manager. If you specify a

2405

handler, it is added to the logger on entry to the block and removed on exit

2406

from the block. You can also ask the manager to close the handler for you on

2407

block exit - you could do this if you don’t need the handler any more.

2408

To illustrate how it works, we can add the following block of code to the

2409

above:

2410

<div class="highlight-python3"><div class="highlight"><pre>if __name__ == '__main__':

2411

2412

2413

2414

logger.info('1. This should appear just once on stderr.')

2415

logger.debug('2. This should not appear.')

2416

2417

logger.debug('3. This should appear once on stderr.')

2418

logger.debug('4. This should not appear.')

2419

h = logging.StreamHandler(sys.stdout)

2420

with LoggingContext(logger, level=logging.DEBUG, handler=h, close=True):

2421

logger.debug('5. This should appear twice - once on stderr and once on stdout.')

2422

logger.info('6. This should appear just once on stderr.')

2423

logger.debug('7. This should not appear.')

2424

</pre></div>

2425

</div>

2426

We initially set the logger’s level to <code class="docutils literal">INFO</code>, so message #1 appears and

2427

message #2 doesn’t. We then change the level to <code class="docutils literal">DEBUG</code> temporarily in the

2428

following <code class="docutils literal">with</code> block, and so message #3 appears. After the block exits, the

2429

logger’s level is restored to <code class="docutils literal">INFO</code> and so message #4 doesn’t appear. In the

2430

next <code class="docutils literal">with</code> block, we set the level to <code class="docutils literal">DEBUG</code> again but also add a handler

2431

writing to <code class="docutils literal">sys.stdout</code>. Thus, message #5 appears twice on the console (once

2432

via <code class="docutils literal">stderr</code> and once via <code class="docutils literal">stdout</code>). After the <code class="docutils literal">with</code> statement’s

2433

completion, the status is as it was before so message #6 appears (like message

2434

#1) whereas message #7 doesn’t (just like message #2).

2435

If we run the resulting script, the result is as follows:

2436

<div class="highlight-shell-session"><div class="highlight"><pre>$ python logctx.py

2437

1. This should appear just once on stderr.

2438

3. This should appear once on stderr.

2439

5. This should appear twice - once on stderr and once on stdout.

2440

5. This should appear twice - once on stderr and once on stdout.

2441

6. This should appear just once on stderr.

2442

</pre></div>

2443

</div>

2444

If we run it again, but pipe <code class="docutils literal">stderr</code> to <code class="docutils literal">/dev/null</code>, we see the following,

2445

which is the only message written to <code class="docutils literal">stdout</code>:

2446

<div class="highlight-shell-session"><div class="highlight"><pre>$ python logctx.py 2>/dev/null

2447

5. This should appear twice - once on stderr and once on stdout.

2448

</pre></div>

2449

</div>

2450

Once again, but piping <code class="docutils literal">stdout</code> to <code class="docutils literal">/dev/null</code>, we get:

2451

<div class="highlight-shell-session"><div class="highlight"><pre>$ python logctx.py >/dev/null

2452

1. This should appear just once on stderr.

2453

3. This should appear once on stderr.

2454

5. This should appear twice - once on stderr and once on stdout.

2455

6. This should appear just once on stderr.

2456

</pre></div>

2457

</div>

2458

In this case, the message #5 printed to <code class="docutils literal">stdout</code> doesn’t appear, as expected.

2459

Of course, the approach described here can be generalised, for example to attach

2460

logging filters temporarily. Note that the above code works in Python 2 as well

2461

as Python 3.

2462

</div>

2463

</div>

2464

2465

2466

</div>

2467

</div>

2468

</div>

2469

2470

2471

<h3><a href="../contents.html">Table Of Contents</a></h3>

2472

<ul>

2473

<li><a class="reference internal" href="#">Logging Cookbook</a><ul>

2474

<li><a class="reference internal" href="#using-logging-in-multiple-modules">Using logging in multiple modules</a></li>

2475

<li><a class="reference internal" href="#logging-from-multiple-threads">Logging from multiple threads</a></li>

2476

<li><a class="reference internal" href="#multiple-handlers-and-formatters">Multiple handlers and formatters</a></li>

2477

<li><a class="reference internal" href="#logging-to-multiple-destinations">Logging to multiple destinations</a></li>

2478

<li><a class="reference internal" href="#configuration-server-example">Configuration server example</a></li>

2479

<li><a class="reference internal" href="#dealing-with-handlers-that-block">Dealing with handlers that block</a></li>

2480

<li><a class="reference internal" href="#sending-and-receiving-logging-events-across-a-network">Sending and receiving logging events across a network</a></li>

2481

<li><a class="reference internal" href="#adding-contextual-information-to-your-logging-output">Adding contextual information to your logging output</a><ul>

2482

<li><a class="reference internal" href="#using-loggeradapters-to-impart-contextual-information">Using LoggerAdapters to impart contextual information</a><ul>

2483

<li><a class="reference internal" href="#using-objects-other-than-dicts-to-pass-contextual-information">Using objects other than dicts to pass contextual information</a></li>

2484

</ul>

2485

</li>

2486

<li><a class="reference internal" href="#using-filters-to-impart-contextual-information">Using Filters to impart contextual information</a></li>

2487

</ul>

2488

</li>

2489

<li><a class="reference internal" href="#logging-to-a-single-file-from-multiple-processes">Logging to a single file from multiple processes</a></li>

2490

<li><a class="reference internal" href="#using-file-rotation">Using file rotation</a></li>

2491

<li><a class="reference internal" href="#use-of-alternative-formatting-styles">Use of alternative formatting styles</a></li>

2492

<li><a class="reference internal" href="#customizing-logrecord">Customizing <code class="docutils literal">LogRecord</code></a></li>

2493

<li><a class="reference internal" href="#subclassing-queuehandler-a-zeromq-example">Subclassing QueueHandler - a ZeroMQ example</a></li>

2494

<li><a class="reference internal" href="#subclassing-queuelistener-a-zeromq-example">Subclassing QueueListener - a ZeroMQ example</a></li>

2495

<li><a class="reference internal" href="#an-example-dictionary-based-configuration">An example dictionary-based configuration</a></li>

2496

<li><a class="reference internal" href="#using-a-rotator-and-namer-to-customize-log-rotation-processing">Using a rotator and namer to customize log rotation processing</a></li>

2497

<li><a class="reference internal" href="#a-more-elaborate-multiprocessing-example">A more elaborate multiprocessing example</a></li>

2498

<li><a class="reference internal" href="#inserting-a-bom-into-messages-sent-to-a-sysloghandler">Inserting a BOM into messages sent to a SysLogHandler</a></li>

2499

<li><a class="reference internal" href="#implementing-structured-logging">Implementing structured logging</a></li>

2500

<li><a class="reference internal" href="#customizing-handlers-with-dictconfig">Customizing handlers with <code class="docutils literal">dictConfig()</code></a></li>

2501

<li><a class="reference internal" href="#using-particular-formatting-styles-throughout-your-application">Using particular formatting styles throughout your application</a><ul>

2502

<li><a class="reference internal" href="#using-logrecord-factories">Using LogRecord factories</a></li>

2503

<li><a class="reference internal" href="#using-custom-message-objects">Using custom message objects</a></li>

2504

</ul>

2505

</li>

2506

<li><a class="reference internal" href="#configuring-filters-with-dictconfig">Configuring filters with <code class="docutils literal">dictConfig()</code></a></li>

2507

<li><a class="reference internal" href="#customized-exception-formatting">Customized exception formatting</a></li>

2508

<li><a class="reference internal" href="#speaking-logging-messages">Speaking logging messages</a></li>

2509

<li><a class="reference internal" href="#buffering-logging-messages-and-outputting-them-conditionally">Buffering logging messages and outputting them conditionally</a></li>

2510

<li><a class="reference internal" href="#formatting-times-using-utc-gmt-via-configuration">Formatting times using UTC (GMT) via configuration</a></li>

2511

<li><a class="reference internal" href="#using-a-context-manager-for-selective-logging">Using a context manager for selective logging</a></li>

2512

</ul>

2513

</li>

2514

</ul>

2515

2516

<h4>Previous topic</h4>

2517

<a href="logging.html"

2518

title="previous chapter">Logging HOWTO</a>

2519

<h4>Next topic</h4>

2520

<a href="regex.html"

2521

title="next chapter">Regular Expression HOWTO</a>

2522

2523

2524

2525

<li><a href="../bugs.html">Report a Bug</a></li>

2526

<li><a href="../_sources/howto/logging-cookbook.txt"

2527

rel="nofollow">Show Source</a></li>

2528

</ul>

2529

</div>

2530

</div>

2531

</div>

2532

2533

</div>

2534

2535

<h3>Navigation</h3>

2536

<ul>

2537

2538

<a href="../genindex.html" title="General Index"

2539

>index</a></li>

2540

2541

<a href="../py-modindex.html" title="Python Module Index"

2542

>modules</a> |</li>

2543

2544

<a href="regex.html" title="Regular Expression HOWTO"

2545

>next</a> |</li>

2546

2547

<a href="logging.html" title="Logging HOWTO"

2548

>previous</a> |</li>

2549

<li><img src="../_static/py.png" alt=""

2550

style="vertical-align: middle; margin-top: -1px"/></li>

2551

<li><a href="https://www.python.org/">Python</a> »</li>

2552

<li>

2553

3.5.2

2554

<a href="../index.html">Documentation </a> »

2555

</li>

2556

2557

<li class="nav-item nav-item-1"><a href="index.html" >Python HOWTOs</a> »</li>

2558

2559

2560

2561

2562

2563

2564

2565

2566

2567

</form>

2568

</div>

2569

2570

2571

</li>

2572

2573

</ul>

2574

</div>

2575

2576

2577

2578

The Python Software Foundation is a non-profit corporation.

2579

<a href="https://www.python.org/psf/donations/">Please donate.</a>

2580

2581

Last updated on Sep 23, 2016.

2582

<a href="../bugs.html">Found a bug</a>?

2583

2584

Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.3.3.

2585

</div>

2586

2587

</body>

2588

</html>

b'\\ No newline at end of file'

Older »