source: rtems/cpukit/zlib/examples/zlib_how.html @ 8198f69

4.11
Last change on this file since 8198f69 was 8198f69, checked in by Ralf Corsepius <ralf.corsepius@…>, on Mar 18, 2011 at 10:10:55 AM

Import zlib-1.2.3

  • Property mode set to 100644
File size: 28.0 KB
Line 
1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"
2  "http://www.w3.org/TR/REC-html40/loose.dtd">
3<html>
4<head>
5<meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
6<title>zlib Usage Example</title>
7<!--  Copyright (c) 2004 Mark Adler.  -->
8</head>
9<body bgcolor="#FFFFFF" text="#000000" link="#0000FF" vlink="#00A000">
10<h2 align="center"> zlib Usage Example </h2>
11We often get questions about how the <tt>deflate()</tt> and <tt>inflate()</tt> functions should be used.
12Users wonder when they should provide more input, when they should use more output,
13what to do with a <tt>Z_BUF_ERROR</tt>, how to make sure the process terminates properly, and
14so on.  So for those who have read <tt>zlib.h</tt> (a few times), and
15would like further edification, below is an annotated example in C of simple routines to compress and decompress
16from an input file to an output file using <tt>deflate()</tt> and <tt>inflate()</tt> respectively.  The
17annotations are interspersed between lines of the code.  So please read between the lines.
18We hope this helps explain some of the intricacies of <em>zlib</em>.
19<p>
20Without further adieu, here is the program <a href="zpipe.c"><tt>zpipe.c</tt></a>:
21<pre><b>
22/* zpipe.c: example of proper use of zlib's inflate() and deflate()
23   Not copyrighted -- provided to the public domain
24   Version 1.2  9 November 2004  Mark Adler */
25
26/* Version history:
27   1.0  30 Oct 2004  First version
28   1.1   8 Nov 2004  Add void casting for unused return values
29                     Use switch statement for inflate() return values
30   1.2   9 Nov 2004  Add assertions to document zlib guarantees
31 */
32</b></pre><!-- -->
33We now include the header files for the required definitions.  From
34<tt>stdio.h</tt> we use <tt>fopen()</tt>, <tt>fread()</tt>, <tt>fwrite()</tt>,
35<tt>feof()</tt>, <tt>ferror()</tt>, and <tt>fclose()</tt> for file i/o, and
36<tt>fputs()</tt> for error messages.  From <tt>string.h</tt> we use
37<tt>strcmp()</tt> for command line argument processing.
38From <tt>assert.h</tt> we use the <tt>assert()</tt> macro.
39From <tt>zlib.h</tt>
40we use the basic compression functions <tt>deflateInit()</tt>,
41<tt>deflate()</tt>, and <tt>deflateEnd()</tt>, and the basic decompression
42functions <tt>inflateInit()</tt>, <tt>inflate()</tt>, and
43<tt>inflateEnd()</tt>.
44<pre><b>
45#include &lt;stdio.h&gt;
46#include &lt;string.h&gt;
47#include &lt;assert.h&gt;
48#include "zlib.h"
49</b></pre><!-- -->
50<tt>CHUNK</tt> is simply the buffer size for feeding data to and pulling data
51from the <em>zlib</em> routines.  Larger buffer sizes would be more efficient,
52especially for <tt>inflate()</tt>.  If the memory is available, buffers sizes
53on the order of 128K or 256K bytes should be used.
54<pre><b>
55#define CHUNK 16384
56</b></pre><!-- -->
57The <tt>def()</tt> routine compresses data from an input file to an output file.  The output data
58will be in the <em>zlib</em> format, which is different from the <em>gzip</em> or <em>zip</em>
59formats.  The <em>zlib</em> format has a very small header of only two bytes to identify it as
60a <em>zlib</em> stream and to provide decoding information, and a four-byte trailer with a fast
61check value to verify the integrity of the uncompressed data after decoding.
62<pre><b>
63/* Compress from file source to file dest until EOF on source.
64   def() returns Z_OK on success, Z_MEM_ERROR if memory could not be
65   allocated for processing, Z_STREAM_ERROR if an invalid compression
66   level is supplied, Z_VERSION_ERROR if the version of zlib.h and the
67   version of the library linked do not match, or Z_ERRNO if there is
68   an error reading or writing the files. */
69int def(FILE *source, FILE *dest, int level)
70{
71</b></pre>
72Here are the local variables for <tt>def()</tt><tt>ret</tt> will be used for <em>zlib</em>
73return codes.  <tt>flush</tt> will keep track of the current flushing state for <tt>deflate()</tt>,
74which is either no flushing, or flush to completion after the end of the input file is reached.
75<tt>have</tt> is the amount of data returned from <tt>deflate()</tt>.  The <tt>strm</tt> structure
76is used to pass information to and from the <em>zlib</em> routines, and to maintain the
77<tt>deflate()</tt> state.  <tt>in</tt> and <tt>out</tt> are the input and output buffers for
78<tt>deflate()</tt>.
79<pre><b>
80    int ret, flush;
81    unsigned have;
82    z_stream strm;
83    char in[CHUNK];
84    char out[CHUNK];
85</b></pre><!-- -->
86The first thing we do is to initialize the <em>zlib</em> state for compression using
87<tt>deflateInit()</tt>.  This must be done before the first use of <tt>deflate()</tt>.
88The <tt>zalloc</tt>, <tt>zfree</tt>, and <tt>opaque</tt> fields in the <tt>strm</tt>
89structure must be initialized before calling <tt>deflateInit()</tt>.  Here they are
90set to the <em>zlib</em> constant <tt>Z_NULL</tt> to request that <em>zlib</em> use
91the default memory allocation routines.  An application may also choose to provide
92custom memory allocation routines here.  <tt>deflateInit()</tt> will allocate on the
93order of 256K bytes for the internal state.
94(See <a href="zlib_tech.html"><em>zlib Technical Details</em></a>.)
95<p>
96<tt>deflateInit()</tt> is called with a pointer to the structure to be initialized and
97the compression level, which is an integer in the range of -1 to 9.  Lower compression
98levels result in faster execution, but less compression.  Higher levels result in
99greater compression, but slower execution.  The <em>zlib</em> constant Z_DEFAULT_COMPRESSION,
100equal to -1,
101provides a good compromise between compression and speed and is equivalent to level 6.
102Level 0 actually does no compression at all, and in fact expands the data slightly to produce
103the <em>zlib</em> format (it is not a byte-for-byte copy of the input).
104More advanced applications of <em>zlib</em>
105may use <tt>deflateInit2()</tt> here instead.  Such an application may want to reduce how
106much memory will be used, at some price in compression.  Or it may need to request a
107<em>gzip</em> header and trailer instead of a <em>zlib</em> header and trailer, or raw
108encoding with no header or trailer at all.
109<p>
110We must check the return value of <tt>deflateInit()</tt> against the <em>zlib</em> constant
111<tt>Z_OK</tt> to make sure that it was able to
112allocate memory for the internal state, and that the provided arguments were valid.
113<tt>deflateInit()</tt> will also check that the version of <em>zlib</em> that the <tt>zlib.h</tt>
114file came from matches the version of <em>zlib</em> actually linked with the program.  This
115is especially important for environments in which <em>zlib</em> is a shared library.
116<p>
117Note that an application can initialize multiple, independent <em>zlib</em> streams, which can
118operate in parallel.  The state information maintained in the structure allows the <em>zlib</em>
119routines to be reentrant.
120<pre><b>
121    /* allocate deflate state */
122    strm.zalloc = Z_NULL;
123    strm.zfree = Z_NULL;
124    strm.opaque = Z_NULL;
125    ret = deflateInit(&amp;strm, level);
126    if (ret != Z_OK)
127        return ret;
128</b></pre><!-- -->
129With the pleasantries out of the way, now we can get down to business.  The outer <tt>do</tt>-loop
130reads all of the input file and exits at the bottom of the loop once end-of-file is reached.
131This loop contains the only call of <tt>deflate()</tt>.  So we must make sure that all of the
132input data has been processed and that all of the output data has been generated and consumed
133before we fall out of the loop at the bottom.
134<pre><b>
135    /* compress until end of file */
136    do {
137</b></pre>
138We start off by reading data from the input file.  The number of bytes read is put directly
139into <tt>avail_in</tt>, and a pointer to those bytes is put into <tt>next_in</tt>.  We also
140check to see if end-of-file on the input has been reached.  If we are at the end of file, then <tt>flush</tt> is set to the
141<em>zlib</em> constant <tt>Z_FINISH</tt>, which is later passed to <tt>deflate()</tt> to
142indicate that this is the last chunk of input data to compress.  We need to use <tt>feof()</tt>
143to check for end-of-file as opposed to seeing if fewer than <tt>CHUNK</tt> bytes have been read.  The
144reason is that if the input file length is an exact multiple of <tt>CHUNK</tt>, we will miss
145the fact that we got to the end-of-file, and not know to tell <tt>deflate()</tt> to finish
146up the compressed stream.  If we are not yet at the end of the input, then the <em>zlib</em>
147constant <tt>Z_NO_FLUSH</tt> will be passed to <tt>deflate</tt> to indicate that we are still
148in the middle of the uncompressed data.
149<p>
150If there is an error in reading from the input file, the process is aborted with
151<tt>deflateEnd()</tt> being called to free the allocated <em>zlib</em> state before returning
152the error.  We wouldn't want a memory leak, now would we?  <tt>deflateEnd()</tt> can be called
153at any time after the state has been initialized.  Once that's done, <tt>deflateInit()</tt> (or
154<tt>deflateInit2()</tt>) would have to be called to start a new compression process.  There is
155no point here in checking the <tt>deflateEnd()</tt> return code.  The deallocation can't fail.
156<pre><b>
157        strm.avail_in = fread(in, 1, CHUNK, source);
158        if (ferror(source)) {
159            (void)deflateEnd(&amp;strm);
160            return Z_ERRNO;
161        }
162        flush = feof(source) ? Z_FINISH : Z_NO_FLUSH;
163        strm.next_in = in;
164</b></pre><!-- -->
165The inner <tt>do</tt>-loop passes our chunk of input data to <tt>deflate()</tt>, and then
166keeps calling <tt>deflate()</tt> until it is done producing output.  Once there is no more
167new output, <tt>deflate()</tt> is guaranteed to have consumed all of the input, i.e.,
168<tt>avail_in</tt> will be zero.
169<pre><b>
170        /* run deflate() on input until output buffer not full, finish
171           compression if all of source has been read in */
172        do {
173</b></pre>
174Output space is provided to <tt>deflate()</tt> by setting <tt>avail_out</tt> to the number
175of available output bytes and <tt>next_out</tt> to a pointer to that space.
176<pre><b>
177            strm.avail_out = CHUNK;
178            strm.next_out = out;
179</b></pre>
180Now we call the compression engine itself, <tt>deflate()</tt>.  It takes as many of the
181<tt>avail_in</tt> bytes at <tt>next_in</tt> as it can process, and writes as many as
182<tt>avail_out</tt> bytes to <tt>next_out</tt>.  Those counters and pointers are then
183updated past the input data consumed and the output data written.  It is the amount of
184output space available that may limit how much input is consumed.
185Hence the inner loop to make sure that
186all of the input is consumed by providing more output space each time.  Since <tt>avail_in</tt>
187and <tt>next_in</tt> are updated by <tt>deflate()</tt>, we don't have to mess with those
188between <tt>deflate()</tt> calls until it's all used up.
189<p>
190The parameters to <tt>deflate()</tt> are a pointer to the <tt>strm</tt> structure containing
191the input and output information and the internal compression engine state, and a parameter
192indicating whether and how to flush data to the output.  Normally <tt>deflate</tt> will consume
193several K bytes of input data before producing any output (except for the header), in order
194to accumulate statistics on the data for optimum compression.  It will then put out a burst of
195compressed data, and proceed to consume more input before the next burst.  Eventually,
196<tt>deflate()</tt>
197must be told to terminate the stream, complete the compression with provided input data, and
198write out the trailer check value.  <tt>deflate()</tt> will continue to compress normally as long
199as the flush parameter is <tt>Z_NO_FLUSH</tt>.  Once the <tt>Z_FINISH</tt> parameter is provided,
200<tt>deflate()</tt> will begin to complete the compressed output stream.  However depending on how
201much output space is provided, <tt>deflate()</tt> may have to be called several times until it
202has provided the complete compressed stream, even after it has consumed all of the input.  The flush
203parameter must continue to be <tt>Z_FINISH</tt> for those subsequent calls.
204<p>
205There are other values of the flush parameter that are used in more advanced applications.  You can
206force <tt>deflate()</tt> to produce a burst of output that encodes all of the input data provided
207so far, even if it wouldn't have otherwise, for example to control data latency on a link with
208compressed data.  You can also ask that <tt>deflate()</tt> do that as well as erase any history up to
209that point so that what follows can be decompressed independently, for example for random access
210applications.  Both requests will degrade compression by an amount depending on how often such
211requests are made.
212<p>
213<tt>deflate()</tt> has a return value that can indicate errors, yet we do not check it here.  Why
214not?  Well, it turns out that <tt>deflate()</tt> can do no wrong here.  Let's go through
215<tt>deflate()</tt>'s return values and dispense with them one by one.  The possible values are
216<tt>Z_OK</tt>, <tt>Z_STREAM_END</tt>, <tt>Z_STREAM_ERROR</tt>, or <tt>Z_BUF_ERROR</tt><tt>Z_OK</tt>
217is, well, ok.  <tt>Z_STREAM_END</tt> is also ok and will be returned for the last call of
218<tt>deflate()</tt>.  This is already guaranteed by calling <tt>deflate()</tt> with <tt>Z_FINISH</tt>
219until it has no more output.  <tt>Z_STREAM_ERROR</tt> is only possible if the stream is not
220initialized properly, but we did initialize it properly.  There is no harm in checking for
221<tt>Z_STREAM_ERROR</tt> here, for example to check for the possibility that some
222other part of the application inadvertently clobbered the memory containing the <em>zlib</em> state.
223<tt>Z_BUF_ERROR</tt> will be explained further below, but
224suffice it to say that this is simply an indication that <tt>deflate()</tt> could not consume
225more input or produce more output.  <tt>deflate()</tt> can be called again with more output space
226or more available input, which it will be in this code.
227<pre><b>
228            ret = deflate(&amp;strm, flush);    /* no bad return value */
229            assert(ret != Z_STREAM_ERROR);  /* state not clobbered */
230</b></pre>
231Now we compute how much output <tt>deflate()</tt> provided on the last call, which is the
232difference between how much space was provided before the call, and how much output space
233is still available after the call.  Then that data, if any, is written to the output file.
234We can then reuse the output buffer for the next call of <tt>deflate()</tt>.  Again if there
235is a file i/o error, we call <tt>deflateEnd()</tt> before returning to avoid a memory leak.
236<pre><b>
237            have = CHUNK - strm.avail_out;
238            if (fwrite(out, 1, have, dest) != have || ferror(dest)) {
239                (void)deflateEnd(&amp;strm);
240                return Z_ERRNO;
241            }
242</b></pre>
243The inner <tt>do</tt>-loop is repeated until the last <tt>deflate()</tt> call fails to fill the
244provided output buffer.  Then we know that <tt>deflate()</tt> has done as much as it can with
245the provided input, and that all of that input has been consumed.  We can then fall out of this
246loop and reuse the input buffer.
247<p>
248The way we tell that <tt>deflate()</tt> has no more output is by seeing that it did not fill
249the output buffer, leaving <tt>avail_out</tt> greater than zero.  However suppose that
250<tt>deflate()</tt> has no more output, but just so happened to exactly fill the output buffer!
251<tt>avail_out</tt> is zero, and we can't tell that <tt>deflate()</tt> has done all it can.
252As far as we know, <tt>deflate()</tt>
253has more output for us.  So we call it again.  But now <tt>deflate()</tt> produces no output
254at all, and <tt>avail_out</tt> remains unchanged as <tt>CHUNK</tt>.  That <tt>deflate()</tt> call
255wasn't able to do anything, either consume input or produce output, and so it returns
256<tt>Z_BUF_ERROR</tt>.  (See, I told you I'd cover this later.)  However this is not a problem at
257all.  Now we finally have the desired indication that <tt>deflate()</tt> is really done,
258and so we drop out of the inner loop to provide more input to <tt>deflate()</tt>.
259<p>
260With <tt>flush</tt> set to <tt>Z_FINISH</tt>, this final set of <tt>deflate()</tt> calls will
261complete the output stream.  Once that is done, subsequent calls of <tt>deflate()</tt> would return
262<tt>Z_STREAM_ERROR</tt> if the flush parameter is not <tt>Z_FINISH</tt>, and do no more processing
263until the state is reinitialized.
264<p>
265Some applications of <em>zlib</em> have two loops that call <tt>deflate()</tt>
266instead of the single inner loop we have here.  The first loop would call
267without flushing and feed all of the data to <tt>deflate()</tt>.  The second loop would call
268<tt>deflate()</tt> with no more
269data and the <tt>Z_FINISH</tt> parameter to complete the process.  As you can see from this
270example, that can be avoided by simply keeping track of the current flush state.
271<pre><b>
272        } while (strm.avail_out == 0);
273        assert(strm.avail_in == 0);     /* all input will be used */
274</b></pre><!-- -->
275Now we check to see if we have already processed all of the input file.  That information was
276saved in the <tt>flush</tt> variable, so we see if that was set to <tt>Z_FINISH</tt>.  If so,
277then we're done and we fall out of the outer loop.  We're guaranteed to get <tt>Z_STREAM_END</tt>
278from the last <tt>deflate()</tt> call, since we ran it until the last chunk of input was
279consumed and all of the output was generated.
280<pre><b>
281        /* done when last data in file processed */
282    } while (flush != Z_FINISH);
283    assert(ret == Z_STREAM_END);        /* stream will be complete */
284</b></pre><!-- -->
285The process is complete, but we still need to deallocate the state to avoid a memory leak
286(or rather more like a memory hemorrhage if you didn't do this).  Then
287finally we can return with a happy return value.
288<pre><b>
289    /* clean up and return */
290    (void)deflateEnd(&amp;strm);
291    return Z_OK;
292}
293</b></pre><!-- -->
294Now we do the same thing for decompression in the <tt>inf()</tt> routine. <tt>inf()</tt>
295decompresses what is hopefully a valid <em>zlib</em> stream from the input file and writes the
296uncompressed data to the output file.  Much of the discussion above for <tt>def()</tt>
297applies to <tt>inf()</tt> as well, so the discussion here will focus on the differences between
298the two.
299<pre><b>
300/* Decompress from file source to file dest until stream ends or EOF.
301   inf() returns Z_OK on success, Z_MEM_ERROR if memory could not be
302   allocated for processing, Z_DATA_ERROR if the deflate data is
303   invalid or incomplete, Z_VERSION_ERROR if the version of zlib.h and
304   the version of the library linked do not match, or Z_ERRNO if there
305   is an error reading or writing the files. */
306int inf(FILE *source, FILE *dest)
307{
308</b></pre>
309The local variables have the same functionality as they do for <tt>def()</tt>.  The
310only difference is that there is no <tt>flush</tt> variable, since <tt>inflate()</tt>
311can tell from the <em>zlib</em> stream itself when the stream is complete.
312<pre><b>
313    int ret;
314    unsigned have;
315    z_stream strm;
316    char in[CHUNK];
317    char out[CHUNK];
318</b></pre><!-- -->
319The initialization of the state is the same, except that there is no compression level,
320of course, and two more elements of the structure are initialized.  <tt>avail_in</tt>
321and <tt>next_in</tt> must be initialized before calling <tt>inflateInit()</tt>.  This
322is because the application has the option to provide the start of the zlib stream in
323order for <tt>inflateInit()</tt> to have access to information about the compression
324method to aid in memory allocation.  In the current implementation of <em>zlib</em>
325(up through versions 1.2.x), the method-dependent memory allocations are deferred to the first call of
326<tt>inflate()</tt> anyway.  However those fields must be initialized since later versions
327of <em>zlib</em> that provide more compression methods may take advantage of this interface.
328In any case, no decompression is performed by <tt>inflateInit()</tt>, so the
329<tt>avail_out</tt> and <tt>next_out</tt> fields do not need to be initialized before calling.
330<p>
331Here <tt>avail_in</tt> is set to zero and <tt>next_in</tt> is set to <tt>Z_NULL</tt> to
332indicate that no input data is being provided.
333<pre><b>
334    /* allocate inflate state */
335    strm.zalloc = Z_NULL;
336    strm.zfree = Z_NULL;
337    strm.opaque = Z_NULL;
338    strm.avail_in = 0;
339    strm.next_in = Z_NULL;
340    ret = inflateInit(&amp;strm);
341    if (ret != Z_OK)
342        return ret;
343</b></pre><!-- -->
344The outer <tt>do</tt>-loop decompresses input until <tt>inflate()</tt> indicates
345that it has reached the end of the compressed data and has produced all of the uncompressed
346output.  This is in contrast to <tt>def()</tt> which processes all of the input file.
347If end-of-file is reached before the compressed data self-terminates, then the compressed
348data is incomplete and an error is returned.
349<pre><b>
350    /* decompress until deflate stream ends or end of file */
351    do {
352</b></pre>
353We read input data and set the <tt>strm</tt> structure accordingly.  If we've reached the
354end of the input file, then we leave the outer loop and report an error, since the
355compressed data is incomplete.  Note that we may read more data than is eventually consumed
356by <tt>inflate()</tt>, if the input file continues past the <em>zlib</em> stream.
357For applications where <em>zlib</em> streams are embedded in other data, this routine would
358need to be modified to return the unused data, or at least indicate how much of the input
359data was not used, so the application would know where to pick up after the <em>zlib</em> stream.
360<pre><b>
361        strm.avail_in = fread(in, 1, CHUNK, source);
362        if (ferror(source)) {
363            (void)inflateEnd(&amp;strm);
364            return Z_ERRNO;
365        }
366        if (strm.avail_in == 0)
367            break;
368        strm.next_in = in;
369</b></pre><!-- -->
370The inner <tt>do</tt>-loop has the same function it did in <tt>def()</tt>, which is to
371keep calling <tt>inflate()</tt> until has generated all of the output it can with the
372provided input.
373<pre><b>
374        /* run inflate() on input until output buffer not full */
375        do {
376</b></pre>
377Just like in <tt>def()</tt>, the same output space is provided for each call of <tt>inflate()</tt>.
378<pre><b>
379            strm.avail_out = CHUNK;
380            strm.next_out = out;
381</b></pre>
382Now we run the decompression engine itself.  There is no need to adjust the flush parameter, since
383the <em>zlib</em> format is self-terminating. The main difference here is that there are
384return values that we need to pay attention to.  <tt>Z_DATA_ERROR</tt>
385indicates that <tt>inflate()</tt> detected an error in the <em>zlib</em> compressed data format,
386which means that either the data is not a <em>zlib</em> stream to begin with, or that the data was
387corrupted somewhere along the way since it was compressed.  The other error to be processed is
388<tt>Z_MEM_ERROR</tt>, which can occur since memory allocation is deferred until <tt>inflate()</tt>
389needs it, unlike <tt>deflate()</tt>, whose memory is allocated at the start by <tt>deflateInit()</tt>.
390<p>
391Advanced applications may use
392<tt>deflateSetDictionary()</tt> to prime <tt>deflate()</tt> with a set of likely data to improve the
393first 32K or so of compression.  This is noted in the <em>zlib</em> header, so <tt>inflate()</tt>
394requests that that dictionary be provided before it can start to decompress.  Without the dictionary,
395correct decompression is not possible.  For this routine, we have no idea what the dictionary is,
396so the <tt>Z_NEED_DICT</tt> indication is converted to a <tt>Z_DATA_ERROR</tt>.
397<p>
398<tt>inflate()</tt> can also return <tt>Z_STREAM_ERROR</tt>, which should not be possible here,
399but could be checked for as noted above for <tt>def()</tt><tt>Z_BUF_ERROR</tt> does not need to be
400checked for here, for the same reasons noted for <tt>def()</tt><tt>Z_STREAM_END</tt> will be
401checked for later.
402<pre><b>
403            ret = inflate(&amp;strm, Z_NO_FLUSH);
404            assert(ret != Z_STREAM_ERROR);  /* state not clobbered */
405            switch (ret) {
406            case Z_NEED_DICT:
407                ret = Z_DATA_ERROR;     /* and fall through */
408            case Z_DATA_ERROR:
409            case Z_MEM_ERROR:
410                (void)inflateEnd(&amp;strm);
411                return ret;
412            }
413</b></pre>
414The output of <tt>inflate()</tt> is handled identically to that of <tt>deflate()</tt>.
415<pre><b>
416            have = CHUNK - strm.avail_out;
417            if (fwrite(out, 1, have, dest) != have || ferror(dest)) {
418                (void)inflateEnd(&amp;strm);
419                return Z_ERRNO;
420            }
421</b></pre>
422The inner <tt>do</tt>-loop ends when <tt>inflate()</tt> has no more output as indicated
423by not filling the output buffer, just as for <tt>deflate()</tt>.  In this case, we cannot
424assert that <tt>strm.avail_in</tt> will be zero, since the deflate stream may end before the file
425does.
426<pre><b>
427        } while (strm.avail_out == 0);
428</b></pre><!-- -->
429The outer <tt>do</tt>-loop ends when <tt>inflate()</tt> reports that it has reached the
430end of the input <em>zlib</em> stream, has completed the decompression and integrity
431check, and has provided all of the output.  This is indicated by the <tt>inflate()</tt>
432return value <tt>Z_STREAM_END</tt>.  The inner loop is guaranteed to leave <tt>ret</tt>
433equal to <tt>Z_STREAM_END</tt> if the last chunk of the input file read contained the end
434of the <em>zlib</em> stream.  So if the return value is not <tt>Z_STREAM_END</tt>, the
435loop continues to read more input.
436<pre><b>
437        /* done when inflate() says it's done */
438    } while (ret != Z_STREAM_END);
439</b></pre><!-- -->
440At this point, decompression successfully completed, or we broke out of the loop due to no
441more data being available from the input file.  If the last <tt>inflate()</tt> return value
442is not <tt>Z_STREAM_END</tt>, then the <em>zlib</em> stream was incomplete and a data error
443is returned.  Otherwise, we return with a happy return value.  Of course, <tt>inflateEnd()</tt>
444is called first to avoid a memory leak.
445<pre><b>
446    /* clean up and return */
447    (void)inflateEnd(&amp;strm);
448    return ret == Z_STREAM_END ? Z_OK : Z_DATA_ERROR;
449}
450</b></pre><!-- -->
451That ends the routines that directly use <em>zlib</em>.  The following routines make this
452a command-line program by running data through the above routines from <tt>stdin</tt> to
453<tt>stdout</tt>, and handling any errors reported by <tt>def()</tt> or <tt>inf()</tt>.
454<p>
455<tt>zerr()</tt> is used to interpret the possible error codes from <tt>def()</tt>
456and <tt>inf()</tt>, as detailed in their comments above, and print out an error message.
457Note that these are only a subset of the possible return values from <tt>deflate()</tt>
458and <tt>inflate()</tt>.
459<pre><b>
460/* report a zlib or i/o error */
461void zerr(int ret)
462{
463    fputs("zpipe: ", stderr);
464    switch (ret) {
465    case Z_ERRNO:
466        if (ferror(stdin))
467            fputs("error reading stdin\n", stderr);
468        if (ferror(stdout))
469            fputs("error writing stdout\n", stderr);
470        break;
471    case Z_STREAM_ERROR:
472        fputs("invalid compression level\n", stderr);
473        break;
474    case Z_DATA_ERROR:
475        fputs("invalid or incomplete deflate data\n", stderr);
476        break;
477    case Z_MEM_ERROR:
478        fputs("out of memory\n", stderr);
479        break;
480    case Z_VERSION_ERROR:
481        fputs("zlib version mismatch!\n", stderr);
482    }
483}
484</b></pre><!-- -->
485Here is the <tt>main()</tt> routine used to test <tt>def()</tt> and <tt>inf()</tt>.  The
486<tt>zpipe</tt> command is simply a compression pipe from <tt>stdin</tt> to <tt>stdout</tt>, if
487no arguments are given, or it is a decompression pipe if <tt>zpipe -d</tt> is used.  If any other
488arguments are provided, no compression or decompression is performed.  Instead a usage
489message is displayed.  Examples are <tt>zpipe < foo.txt > foo.txt.z</tt> to compress, and
490<tt>zpipe -d < foo.txt.z > foo.txt</tt> to decompress.
491<pre><b>
492/* compress or decompress from stdin to stdout */
493int main(int argc, char **argv)
494{
495    int ret;
496
497    /* do compression if no arguments */
498    if (argc == 1) {
499        ret = def(stdin, stdout, Z_DEFAULT_COMPRESSION);
500        if (ret != Z_OK)
501            zerr(ret);
502        return ret;
503    }
504
505    /* do decompression if -d specified */
506    else if (argc == 2 &amp;&amp; strcmp(argv[1], "-d") == 0) {
507        ret = inf(stdin, stdout);
508        if (ret != Z_OK)
509            zerr(ret);
510        return ret;
511    }
512
513    /* otherwise, report usage */
514    else {
515        fputs("zpipe usage: zpipe [-d] &lt; source &gt; dest\n", stderr);
516        return 1;
517    }
518}
519</b></pre>
520<hr>
521<i>Copyright (c) 2004 by Mark Adler<br>Last modified 13 November 2004</i>
522</body>
523</html>
Note: See TracBrowser for help on using the repository browser.