zzip-file.htm

<section> <date> 1. June 2000 </date>
<h2> ZIP File Access </h2>  Using Zipped Files Transparently 

<!--border-->

<section>
<h3>The Typedef</h3>

<P>
 The typedef <code>ZZIP_FILE</code> can serve as a replacement 
 for a normal file descriptor. As long as it is only used
 for reading a file, the zzlib-user can actually replace
 the posix functions <code>open/read/close</code> 
 by their counterparts from the 
 <a href="zziplib.html">zziplib library</a>:
 <code>zzip_open/zzip_read/zzip_close</code>.
</P>
<P>
 As long as the filename path given to <code>zzip_open</code>
 refers to a real file in the filesystem, it will almost
 directly forward the call to the respective posix <code>open</code>
 call. The returned file descriptor is then stored in
 a member-variable of the <code>ZZIP_FILE</code> structure.
</P>
<P>
 Any subsequent calls to <code>zzip_read</code> will then
 be forwarded to the posix <code>read</code> call on the
 memorized file descriptor. The same about <code>zzip_close</code>
 which will call the posix <code>close</code> function and then
 <code>free</code> the <code>ZZIP_FILE</code> structure.
</P>
<P>
 The real benefit of the 
 <a href="zziplib.html">zziplib library</a>
 comes about when the filename argument does actually refer
 to a file that is zipped in a zip-archive. It happens that
 even both a real file and a zipped file can live under the
 same pathname given to the <code>zzip_open</code> call,
 whereas the real file is used in preference.
</P>

</section><section>
<h3>Zipped File</h3>

<P>
 Suppose you have subdirectory called '<tt>test/</tt>'. In
 this directory is just one file, called '<tt>README</tt>'.
 Calling the <code>zzip_open</code> function with an
 argument of '<i>optional-path/</i> <tt>test/README</tt>',
 then it will open that file for subsequent reading with
 <code>zzip_read</code>. In this case the real (<i>stat'able</i>)
 file is opened.
</P>
<P>
 Now you can go to the '<tt>test/</tt>' directory and zip up
 the files in there by calling 
 <nobr><tt>`zip ../test.zip *`</tt></nobr>.
 After this, you can delete the '<tt>test/</tt>' directory and
 the call to <code>zzip_open</code> will still succeed.
 The reason is that the part of the path saying 
 '<tt>test/README</tt>' will be replaced by sth. like 
 '<tt>test.zip:README</tt>' - that is the real file '<tt>test.zip</tt>'
 is opened and searched for a contained file '<tt>README</tt>'.
</P>
<P>
 Calling <code>zzip_read</code> on the zipped '<tt>README</tt>' file
 will return the very same data as if it is a real file in a
 real directory. If the zipped file is compressed it will be 
 decompressed on the fly.
</P>

</section><section>
<h3>Zip Directory</h3>

<P>
 The same applies to the use of <code>opendir/readdir/closedir</code>
 which can safely be replaced with their counterparts from the
 <a href="zziplib.html">zziplib library</a> - again their prototype
 follows the scheme of the original calls, just prepend <tt>zzip_</tt>
 to the function calls and <tt>ZZIP_</tt> to the struct-typedefs.
</P>
<P>
 To call <code>zzip_opendir</code> on a real directory will then
 return a <code>ZZIP_DIR</code> whose member-variable 
 <code>realdir</code> points to the actual <code>DIR</code>-structure
 returned by the underlying posix <code>opendir</code>-call.
</P>
<P>
 If a real directory '<tt>test</tt>' does not exist, then the
 <code>zzip_opendir</code> will try to open a file '<tt>test.zip</tt>'
 with a call to <code>zzip_dir_open</code>.
 Subsequent calls to <code>zzip_readdir</code> will then return
 information as being obtained from the central archive directory
 of the zip-file.
</P>

</section><section>
<h3>Differences</h3>

<P>
 There are no differences between the posix calls and their counterparts
 from the      <a href="zziplib.html">zziplib library</a> - well, just
 as long as the zip-file contains just the plain files from a directory.
</P>
<P>
 If the zip-file contains directory entries you may be prompted with
 some awkward behaviour, since in zip-file a directory happens to be
 just an empty file. Note that the posix function <code>open</code>
 may also open a directory for reading - it will only return 
 <code>EISDIR</code> if the <code>open</code> mode-argument included
 write-access.
</P>
<P>
 What the current of version of the 
 <a href="zziplib.html">zziplib library</a>
 can definitely not do: calling zzip_opendir on a directory zippend
 <em>inside</em> a zip-file.
</P>
<P>
 To prevent the enrollment of directories into the zip-archive, you
 can use the <tt>-D</tt> option of the <tt>zip</tt> program. That
 is in any <tt>Makefile</tt> you may want to use
 <nobr><tt>`cd $(dir) &amp;&amp; zip -D ../$(dir).zip *`</tt></nobr>.
</P> 

</section><section>
<h3>Advantages</h3>

<P>
 Distribution of a set of files is much easier if it just means
 to wrap up a group of files into a zip-archive - and copy that
 zip-archive to the respective destination directory.
 Even more the files can be compressed and unlike a <tt>tar.gz</tt>
 archive there is no need to decompress the archive in temporary
 location before accessing a member-file.
</P>
<P>
 On the other hand, there is no chance to scatter files around
 on the disk like it could easily happen with a set of gzip'ed
 man-pages in a single `<tt>man</tt>`-directory. The reader
 application does not specifically need to know that the file
 is compressed, so that reading a script like 
 `<tt>share/guile/x.x.x/ice-9/popen.scm</tt>` is done by simple
 calls to <code>zzip_read</code> which works on zip-file named
 `<tt>share/guile/x.x.x/ice-9.zip</tt>`.
</P>
<P>
 A version mismatch between different files in a group is now
 obvious: either the opened file belongs to the distribution
 archive, or otherwise in resides in a real directory <em>just
 next to the zip-archive that contains the original</em>.
</P>

</section><section>
<h3>Issues</h3>

<P>
 The  <a href="zziplib.html">zziplib library</a> does not
 use any code piece from the <code>zip</code> programs, neither
 <em>pkzip</em> nor <em>infozip</em>, so there is no license
 issue here. The decompression is done by using the free
 <a href="http://www.gzip.org/zlib">zlib library</a> which has no special
 issues with respect to licensing. 
 The  rights to the <a href="zziplib.html">zziplib library</a> 
 are reserved to the copyright holders, there is a public
 license that puts most the sources themselves under 
 <a href="COPYING.LIB">the GNU Lesser General Public License</a>,
 so that the use of a shared library instance of the
 <a href="zziplib.html">zziplib library</a>
 has no restrictions of interest to application programmers.
 For more details and hints about static linking, check
 the <a href="copying.html">COPYING</a> information.
</P>
<P>
 The only issue you have with the
 <a href="zziplib.html">zziplib library</a>
 is the fact that you can only <em>read</em> the contained files.
 Writing/Compression is not implemented. Even more, a compressed
 file is not seekable at the moment although I hope that someone
 will stand up to implement that functionality someday.
</P>

</section></section>