1
<section> <date> 1. June 2000 </date>
2
<h2> ZIP File Access </h2> Using Zipped Files Transparently
10
The typedef <code>ZZIP_FILE</code> can serve as a replacement
11
for a normal file descriptor. As long as it is only used
12
for reading a file, the zzlib-user can actually replace
13
the posix functions <code>open/read/close</code>
14
by their counterparts from the
15
<a href="zziplib.html">zziplib library</a>:
16
<code>zzip_open/zzip_read/zzip_close</code>.
19
As long as the filename path given to <code>zzip_open</code>
20
refers to a real file in the filesystem, it will almost
21
directly forward the call to the respective posix <code>open</code>
22
call. The returned file descriptor is then stored in
23
a member-variable of the <code>ZZIP_FILE</code> structure.
26
Any subsequent calls to <code>zzip_read</code> will then
27
be forwarded to the posix <code>read</code> call on the
28
memorized file descriptor. The same about <code>zzip_close</code>
29
which will call the posix <code>close</code> function and then
30
<code>free</code> the <code>ZZIP_FILE</code> structure.
33
The real benefit of the
34
<a href="zziplib.html">zziplib library</a>
35
comes about when the filename argument does actually refer
36
to a file that is zipped in a zip-archive. It happens that
37
even both a real file and a zipped file can live under the
38
same pathname given to the <code>zzip_open</code> call,
39
whereas the real file is used in preference.
46
Suppose you have subdirectory called '<tt>test/</tt>'. In
47
this directory is just one file, called '<tt>README</tt>'.
48
Calling the <code>zzip_open</code> function with an
49
argument of '<i>optional-path/</i> <tt>test/README</tt>',
50
then it will open that file for subsequent reading with
51
<code>zzip_read</code>. In this case the real (<i>stat'able</i>)
55
Now you can go to the '<tt>test/</tt>' directory and zip up
56
the files in there by calling
57
<nobr><tt>`zip ../test.zip *`</tt></nobr>.
58
After this, you can delete the '<tt>test/</tt>' directory and
59
the call to <code>zzip_open</code> will still succeed.
60
The reason is that the part of the path saying
61
'<tt>test/README</tt>' will be replaced by sth. like
62
'<tt>test.zip:README</tt>' - that is the real file '<tt>test.zip</tt>'
63
is opened and searched for a contained file '<tt>README</tt>'.
66
Calling <code>zzip_read</code> on the zipped '<tt>README</tt>' file
67
will return the very same data as if it is a real file in a
68
real directory. If the zipped file is compressed it will be
69
decompressed on the fly.
73
<h3>Zip Directory</h3>
76
The same applies to the use of <code>opendir/readdir/closedir</code>
77
which can safely be replaced with their counterparts from the
78
<a href="zziplib.html">zziplib library</a> - again their prototype
79
follows the scheme of the original calls, just prepend <tt>zzip_</tt>
80
to the function calls and <tt>ZZIP_</tt> to the struct-typedefs.
83
To call <code>zzip_opendir</code> on a real directory will then
84
return a <code>ZZIP_DIR</code> whose member-variable
85
<code>realdir</code> points to the actual <code>DIR</code>-structure
86
returned by the underlying posix <code>opendir</code>-call.
89
If a real directory '<tt>test</tt>' does not exist, then the
90
<code>zzip_opendir</code> will try to open a file '<tt>test.zip</tt>'
91
with a call to <code>zzip_dir_open</code>.
92
Subsequent calls to <code>zzip_readdir</code> will then return
93
information as being obtained from the central archive directory
101
There are no differences between the posix calls and their counterparts
102
from the <a href="zziplib.html">zziplib library</a> - well, just
103
as long as the zip-file contains just the plain files from a directory.
106
If the zip-file contains directory entries you may be prompted with
107
some awkward behaviour, since in zip-file a directory happens to be
108
just an empty file. Note that the posix function <code>open</code>
109
may also open a directory for reading - it will only return
110
<code>EISDIR</code> if the <code>open</code> mode-argument included
114
What the current of version of the
115
<a href="zziplib.html">zziplib library</a>
116
can definitly not do: calling zzip_opendir on a directory zippend
117
<em>inside</em> a zip-file.
120
To prevent the enrollment of directories into the zip-archive, you
121
can use the <tt>-D</tt> option of the <tt>zip</tt> program. That
122
is in any <tt>Makefile</tt> you may want to use
123
<nobr><tt>`cd $(dir) && zip -D ../$(dir).zip *`</tt></nobr>.
130
Distribution of a set of files is much easier if it just means
131
to wrap up a group of files into a zip-archive - and copy that
132
zip-archive to the respective destination directory.
133
Even more the files can be compressed and unlike a <tt>tar.gz</tt>
134
archive there is no need to decompress the archive in temporary
135
location before accessing a member-file.
138
On the other hand, there is no chance to scatter files around
139
on the disk like it could easily happen with a set of gzip'ed
140
man-pages in a single `<tt>man</tt>`-directory. The reader
141
application does not specifically need to know that the file
142
is compressed, so that reading a script like
143
`<tt>share/guile/x.x.x/ice-9/popen.scm</tt>` is done by simple
144
calls to <code>zzip_read</code> which works on zip-file named
145
`<tt>share/guile/x.x.x/ice-9.zip</tt>`.
148
A version mismatch between different files in a group is now
149
obvious: either the opened file belongs to the distribution
150
archive, or otherwise in resides in a real directory <em>just
151
next to the zip-archive that contains the original</em>.
158
The <a href="zziplib.html">zziplib library</a> does not
159
use any code piece from the <code>zip</code> programs, neither
160
<em>pkzip</em> nor <em>infozip</em>, so there is no license
161
issue here. The decompression is done by using the free
162
<a href="http://www.gzip.org/zlib">zlib library</a> which has no special
163
issues with respect to licensing.
164
The rights to the <a href="zziplib.html">zziplib library</a>
165
are reserved to the copyright holders, there is a public
166
license that puts most the sources themselves under
167
<a href="COPYING.LIB">the GNU Lesser General Public License</a>,
168
so that the use of a shared library instance of the
169
<a href="zziplib.html">zziplib library</a>
170
has no restrictions of interest to application programmers.
171
For more details and hints about static linking, check
172
the <a href="copying.html">COPYING</a> information.
175
The only issue you have with the
176
<a href="zziplib.html">zziplib library</a>
177
is the fact that you can only <em>read</em> the contained files.
178
Writing/Compression is not implemented. Even more, a compressed
179
file is not seekable at the moment although I hope that someone
180
will stand up to implement that functionality someday.