Did some simple benchmarking of gzcompress() this morning on a lightly-loaded Fedora 12 server with an AMD Phenom II 940 (quad-core) rocking 8 gigs of ram and executing PHP 5.3.2.2 as an Apache module:
Compression benchmark: (level, time, size (%)):
0: 0.000373 - 82.08 kB (100.02%)
1: 0.000914 - 19.61 kB (23.90%)
2: 0.000951 - 18.88 kB (23.01%)
3: 0.000999 - 18.43 kB (22.46%)
4: 0.001498 - 17.65 kB (21.51%)
5: 0.001744 - 17.09 kB (20.82%)
6: 0.002060 - 16.88 kB (20.57%)
7: 0.002233 - 16.85 kB (20.53%)
8: 0.002808 - 16.71 kB (20.36%)
9: 0.002928 - 16.71 kB (20.36%)
The time code evaluates to:
1. Get start microtime
2. Call gzcompress
3. Get end microtime
4. Report average duration of 100 cycles
The 82.08 kB was a copied and pasted string of two actual, PHP-generated pages we use on our intranet. Running some rough calculations, the time it takes to compress the data at level 9 will never be larger than the time it takes to transmit slightly-less compressed data at levels 6 or lower. In other words, in our application, we have no reason not to fully-compress each PHP script's output.
PHP - Manual: gzcompress
2024-11-15
gzcompress
(PHP 4 >= 4.0.1, PHP 5, PHP 7, PHP 8)
gzcompress — Compress a string
说明
gzcompress(string
$data, int $level = -1, int $encoding = ZLIB_ENCODING_DEFLATE): string|false
This function compresses the given string using the ZLIB
data format.
For details on the ZLIB compression algorithm see the document "» ZLIB Compressed Data Format Specification version 3.3" (RFC 1950).
注意:
This is not the same as gzip compression, which includes some header data. See gzencode() for gzip compression.
参数
-
data -
The data to compress.
-
level -
The level of compression. Can be given as 0 for no compression up to 9 for maximum compression.
If -1 is used, the default compression of the zlib library is used which is 6.
-
encoding -
One of
ZLIB_ENCODING_*constants.
返回值
The compressed string or false if an error occurred.
范例
示例 #1 gzcompress() example
<?php
$compressed = gzcompress('Compress me', 9);
echo $compressed;
?>
参见
- gzdeflate() - Deflate a string
- gzinflate() - Inflate a deflated string
- gzuncompress() - Uncompress a compressed string
- gzencode() - Create a gzip compressed string
add a note
User Contributed Notes 24 notes
Eric ¶
10 years ago
fil at rezox dot com ¶
21 years ago
GZ-compression works perfectly for html content sent to IE5/IE5.5/Netscape6 browsers. I even compress the <script src='javascriptstuff'> javascript code. It speeds throughput immensely: best case so far 160k html file --> 8k gzip sent.
Netscape 4 (on all platforms[?]), IE4/IE5 on the Mac do NOT support gz-compression. Email me if you want to know more.
ck at newsclub dot de ¶
21 years ago
Ah right, I did not realize it because my browsers (Acorn Browse on RISC OS and Netscape on Windows) don't make problems at all.
Anyway, you can replace your function "gzip_PrintFourChars"
with "echo pack("V",$Crc)".
@boas.anthro.mnsu.edu ¶
21 years ago
No, it doesn't return gzip compressed data -- specifically, the CRC is messed up. However, after massaging the output a lot, I have come up with a solution. I also commented it a lot, pointing out odd things.
<?php
// Start the output buffer
ob_start();
ob_implicit_flush(0);
// Output stuff here...
// Get the contents of the output buffer
$contents = ob_get_contents();
ob_end_clean();
// Tell the browser that they are going to get gzip data
// Of course, you already checked if they support gzip or x-gzip
// and if they support x-gzip, you'd change the header to say
// x-gzip instead, right?
header("Content-Encoding: gzip");
// Display the header of the gzip file
// Thanks ck@medienkombinat.de!
// Only display this once
echo "\x1f\x8b\x08\x00\x00\x00\x00\x00";
// Figure out the size and CRC of the original for later
$Size = strlen($contents);
$Crc = crc32($contents);
// Compress the data
$contents = gzcompress($contents, 9);
// We can't just output it here, since the CRC is messed up.
// If I try to "echo $contents" at this point, the compressed
// data is sent, but not completely. There are four bytes at
// the end that are a CRC. Three are sent. The last one is
// left in limbo. Also, if we "echo $contents", then the next
// byte we echo will not be sent to the client. I am not sure
// if this is a bug in 4.0.2 or not, but the best way to avoid
// this is to put the correct CRC at the end of the compressed
// data. (The one generated by gzcompress looks WAY wrong.)
// This will stop Opera from crashing, gunzip will work, and
// other browsers won't keep loading indefinately.
//
// Strip off the old CRC (it's there, but it won't be displayed
// all the way -- very odd)
$contents = substr($contents, 0, strlen($contents) - 4);
// Show only the compressed data
echo $contents;
// Output the CRC, then the size of the original
gzip_PrintFourChars($Crc);
gzip_PrintFourChars($Size);
// Done. You can append further data by gzcompressing
// another string and reworking the CRC and Size stuff for
// it too. Repeat until done.
function gzip_PrintFourChars($Val)
{
for ($i = 0; $i < 4; $i ++)
{
echo chr($Val % 256);
$Val = floor($Val / 256);
}
}
?>
gvirgo at mithril dot com ¶
21 years ago
>>Netscape 4 (on all platforms[?]), IE4/IE5 on the Mac do NOT support
>>gz-compression. Email me if you want to know more.
As a matter of fact, Netscape 4.7x on Win32 DOES support gz-compression. Calling the following script in my auto_prepend directive seems to work fine. The script is taken from the Output Buffering article on Zend.com (http://www.zend.com/zend/art/buffering.php). As noted in the article, it only works with PHP 4.0.4 and higher.
<?php
// function to compress the output buffer in preparation for transmission to client
function compress_output($output) {
// compress and return the output buffer
return gzencode($output);
}
// Check if the browser supports gzip encoding, HTTP_ACCEPT_ENCODING
if (strstr($HTTP_SERVER_VARS['HTTP_ACCEPT_ENCODING'], 'gzip')) {
// Start output buffering, and register compress_output()
ob_start("compress_output");
// Tell the browser the content is compressed with gzip
header("Content-Encoding: gzip");
}
?>
Anonymous ¶
17 years ago
If no parameter is given, the default level of 6 is used.
benchmarking the speed:
(compression, datasize, compress time)
0: 37440 (0.002 s)
1: 2537 (0.002 s)
2: 2298 (0.003 s)
3: 2199 (0.003 s)
4: 2299 (0.003 s)
5: 1939 (0.004 s)
6: 1789 (0.004 s)
7: 1724 (0.004 s)
8: 1635 (0.005 s)
9: 1635 (0.005 s)
Level 0 = no compression, then level 1-4 are roughly the fastest to compress with reasonable quality. level 5-7 give better compression but take a little longer and 8-9 take nearly double the time to compress but give best results.
So the system default of 6 is actually a good compromise in speed/size. But for very fast compression times, I go with the threshold value of 3.
detain at interserver dot net ¶
4 years ago
gzipped strings include header/metadata you can use to determine if a string is gzipped or not , but since gzcompress does not include that I found myself needing a way to determine if a string was compressed or not. After some research (and then improvements) i came up with this:
/**
* determines if a string is a gzipped string supporting strings
* encoded with either gzencode or gzcompress
*
* @param string $string the string to check for compression
* @return bool whether or not the string was compmressed
*/
function is_gzipped($string) {
return mb_strpos($string, "\x1f\x8b\x08", 'US-ASCII') === 0 && @gzuncompress($string) !== FALSE;
}
Samnan ¶
13 years ago
To properly use gzcompress to send compressed data to browsers that support it (most of modern browsers do, even mobile ones !), use this function (I dont think W3C validator issue mentioned earlier is valid ). use ob_start() at start of your script also
<?php
function print_gzipped_output()
{
$HTTP_ACCEPT_ENCODING = $_SERVER["HTTP_ACCEPT_ENCODING"];
if( headers_sent() )
$encoding = false;
else if( strpos($HTTP_ACCEPT_ENCODING, 'x-gzip') !== false )
$encoding = 'x-gzip';
else if( strpos($HTTP_ACCEPT_ENCODING,'gzip') !== false )
$encoding = 'gzip';
else
$encoding = false;
if( $encoding )
{
$contents = ob_get_clean();
$_temp1 = strlen($contents);
if ($_temp1 < 2048) // no need to waste resources in compressing very little data
print($contents);
else
{
header('Content-Encoding: '.$encoding);
print("\x1f\x8b\x08\x00\x00\x00\x00\x00");
$contents = gzcompress($contents, 9);
$contents = substr($contents, 0, $_temp1);
print($contents);
}
}
else
ob_end_flush();
}
?>
rostor at swzone dot org ¶
20 years ago
To solve I.E.5.x data compress problem on IIS5, such opening a download dial box or wrong page load, the only way for me was to change the header content type:
<? header("Content-type: gzip");ob_start ("ob_gzhandler");?>
but now all browser that doesn't support gzip don't load the page.
webmaster at mail.bg ¶
20 years ago
When you compress your data IE (5.x for sure, no idea about Netscape) may mess up the mime headers, so you may e.g. loose your extensions after compressing output. In my case in attachment saving with no compression everything works fine. When I compress output everything seems OK except I loose the file extension (althogh it really is in the headers). This is because IE seems to think your attachment is mimetype gzip, and not what you send in the headers that follow.
In such cases the easiest sollution would be to disable compression (maybe partially with an .htaccess file?).
devcontact at tech-island dot com ¶
19 years ago
The method of first reading the source file and then passing its content to the gzip function instead of simply the source and destination filename was a bit confusing for me.
So I have written a simple funtion you can use to compress files in the gzip format (gzip is readable by winzip like .zip files)
function compress($srcName, $dstName)
{
$fp = fopen($srcName, "r");
$data = fread ($fp, filesize($srcName));
fclose($fp);
$zp = gzopen($dstName, "w9");
gzwrite($zp, $data);
gzclose($zp);
}
// Compress a file
compress("/web/myfile.dat", "/web/myfile.gz");
plasma ¶
13 years ago
Here is a benchmark comparing strengths of bzip and gzip compression:
gz[0] - 1.647ms - 189058 Byte (100%)
gz[1] - 3.333ms - 44572 Byte (23.58%)
gz[2] - 3.519ms - 42685 Byte (22.58%)
gz[3] - 4.713ms - 40840 Byte (21.6%)
gz[4] - 5.274ms - 39111 Byte (20.69%)
gz[5] - 6.526ms - 37039 Byte (19.59%)
gz[6] - 8.036ms - 36206 Byte (19.15%)
gz[7] - 9.098ms - 35940 Byte (19.01%)
gz[8] - 12.87ms - 35713 Byte (18.89%)
gz[9] - 14.319ms - 35707 Byte (18.89%)
bz[1] - 40.282ms - 32247 Byte (17.06%)
bz[2] - 43.716ms - 29690 Byte (15.7%)
bz[3] - 43.765ms - 29690 Byte (15.7%)
bz[4] - 44.213ms - 29690 Byte (15.7%)
bz[5] - 43.704ms - 29690 Byte (15.7%)
bz[6] - 43.814ms - 29690 Byte (15.7%)
bz[7] - 43.934ms - 29690 Byte (15.7%)
bz[8] - 43.725ms - 29690 Byte (15.7%)
bz[9] - 45.129ms - 29690 Byte (15.7%)
stockton at wowway dot com ¶
8 years ago
After some searching and experimentation I found that output generated with gzcompress() can be uncompressed by the objective c 'zlibInflate' wrapper for 'zlib' as available at http://cocoadev.com/wiki/NSDataCategory but that 'gzdeflate()' cannot. I hope this saves someone else the searching.
Timo ¶
13 years ago
With PHP it is quite easy to modify SWF files. The compression algorithm of SWF is Zlib.
SWF default compression level is 6, at least Flash CS2 uses this level. I tested this by uncompressing compressed SWF file and then compressed the contents with levels 0-9. With level 6, the contents of files were identical.
Maybe it can be whatever else, but when trying to compare files at hex level or compare filesizes, it is easier to use the same level.
prgTW at poczta dot onet dot pl ¶
16 years ago
I don't know why, but W3C Validator can't validate pages that are GZIP compressed (with functions posted below: gzcompress, pack etc.; not by ob_gzhandler).
Because of this, I use similar code to this below:
<?php
if( strstr($_SERVER['HTTP_USER_AGENT'], 'W3C_Validator')!==false || strstr($_SERVER['HTTP_ACCEPT_ENCODING'], 'gzip')===false ){
// do not compress
}else{
// can send GZIP compressed data
}
?>
This code also checks if user agent is W3C Validator. If it is NOT W3C Validator and user acceptes GZIP encoded data then we can compress the page, in other case if it is W3C Validator or user agent does not accept GZIP enoded data we can't compress the page, 'cause user agent couldn't properly handle it.
Sorry for my bad English
Chris Watt ¶
16 years ago
Contrary to popular opinion the checksum in the last four bytes of output is not "wrong", it's just not very useful: RFC 1950 s8.2 specifies that the "Adler-32" checksum, not CRC32 should be appended to the compressed data and PHP function follows the RFC (if for no other reason because it uses zlib and Mark Adler was one of the authors thereof). In contrast gzip expects a CRC32 (i.e. it doesn't follow that RFC), and treats the Adler-32 checksum as an incorrect CRC32.
AFAICT there is no simple way to compute the CRC32 directly from the Adler32 checksum.
Personally, my solution was to write a short C program that performs a deflate on, and calculates the CRC32 of, stdin while writing the compressed data (with trailing CRC32 instead of Adler32) on stdout. It's an ugly hack, but it does allow me to serve a ~1GB compressed output with only a single pass over the uncompressed data and without hitting my 8MB memory limit. In contrast, methods using gzcompress() and crc32() would seem to require a theoretical minimum of three passes across the input data (read into memory, compress, calculate CRC) and 2GB of allocated memory to produce a 1GB compressed output... Short of writing crc32() and deflate() functions from scratch, I couldn't find any way of doing this without either storing the whole output in a variable or resorting to a module in another language.
dwales at easyetrader dot com ¶
18 years ago
I find this works well, save the snip below in your php/ directory then edit your prepend.php file and add a require statement that requires the compression.php file below. Once complete watch (tail -f) your php error log to see the results of compression in action.
<?php
// Gzip encode the contents of the output buffer.
function compress_output_option($output)
{
// Compress the data into a new var.
$compressed_out = gzencode($output);
// Don't compress any pages less than 1000 bytes
// as it's not worth the overhead at either side.
if(strlen($output) >= 1000)
{
error_log('compression.php Gzipd Output'."\n"
.'Before compression size '
.strlen($output).' bytes'."\n"
.' After compression size '
.strlen($compressed_out).' bytes');
// Tell the browser the content is compressed with gzip
header("Content-Encoding: gzip");
return $compressed_out;
}
else
{
// No compression
error_log('compression.php Standard Output.');
return $output;
}
}
// Check if the browser supports gzip encoding, HTTP_ACCEPT_ENCODING
if (strstr($HTTP_SERVER_VARS['HTTP_ACCEPT_ENCODING'], 'gzip'))
{
// Start output buffering, and register compress_output() (see
// below)
ob_start("compress_output_option");
}
?>
mail at ciruz dot de ¶
19 years ago
You should use the crc coz' some linux browsers need it. Some browser won't display anything without crc. My code works with these browsers (tested ;))
gaelreth at gmail dot com ¶
17 years ago
There is a useful zipfile class at http://www.zend.com/codex.php?id=696&single=1 that allows you to zip many files into one zip file.
postmaster at ak-dev dot com ¶
20 years ago
Second try! In my mind, the following source code is right with 4.1.2 and higher, and seems to be right with other versions... and its works with all browser I tried(NS 4.7x, mozilla 1.00rc3, IE5.5, opera6.01)...
As it is said in documentation of the function gzencode, you have to use this function, if you want to get the right result! gzcompress doesn't create the right stream output for browsers!(It is explained in gzencode documentation: generated headers are not the same!)
The source code:
<?
function checkCanGzip() {
global $HTTP_ACCEPT_ENCODING;
if (headers_sent()) return 0;
if (strpos($HTTP_ACCEPT_ENCODING, 'x-gzip') !== false) return
"x-gzip";
if (strpos($HTTP_ACCEPT_ENCODING,'gzip') !== false) return
"gzip";
return 0;
}
function gzDocOut() {
if ($encoding = checkCanGzip()) {
$contents = ob_get_contents();
ob_end_clean();
header("Content-Encoding: ".$encoding);
// print("\x1f\x8b\x08\x00\x00\x00\x00\x00");
$contents = gzencode($contents);
// $contents = substr($contents, 0, strlen($contents) - 4);
print($contents);
// print(pack('V', crc32($contents)));
// print(pack('V', strlen($contents)));
exit();
}
else {
ob_end_flush();
exit();
}
}
ob_start();
ob_implicit_flush(0);
print("your stuff...");
gzDocOut();
?>
You have to note, that gzencode is able to generate deflate compression, but the previous source code, does not implemente it. You just have to read gzencode documentation to find this tip.
function gzencode in php 4.2.0 (and higher) allows to use a compression level, it is not possible in loder versions.
The code which is commented, is wrong source code, used, in the different exemple found on this page!
chris at mad-teaparty dot com ¶
19 years ago
It seems that you don't have to pack the crc-checksum and the length of the string. The crucial thing is to erase the last 4 bytes of the gzcompressed content. They seem to be redundant in any case. I rewrote the function "gzDocOut" and I think, it is the most efficient and shortest way to encode an output with gzcompress:
function gzDocOut() {
if ($encoding = checkCanGzip()) {
$contents = ob_get_contents();
ob_end_clean();
header("Content-Encoding: ". $encoding);
$output = "\x1f\x8b\x08\x00\x00\x00\x00\x00";
$output .= substr(gzcompress($contents, 2), 0, -4);
echo $output;
exit();
}
else {
ob_end_flush();
exit();
}
}
Tested with IE5.5, Netscape 7, Mozilla 1.2, Opera 7, Opera 6 as well as Netscape 4.7. I'll test further browsers to be sure...
orlink at hotmail dot com ¶
20 years ago
I tried sending a pre-compressed javascript with header("Content-Encoding: gzip") to a request from <script type=javascript src="http://...?op=getscript">.
I have found out so far that IE4 does not support external compressed scripts and stylesheets (that is not embedded in the HTML but included with <script ...src=...> and <link src=...>). With other versions of IE it works on some computers and not on others even if they have exactly the same versions of IE and Windows. For instance IE 5.0xxxxxx and Windows 98 Buldxxxx.
If anyones knows of at least one configuration of IE and a OS for which it works for SURE let me know.
chris at mad-teaparty dot com ¶
19 years ago
little correction of the one above:
the crc is not necessarily needed to display the content in browsers, but it's still useful as a checksum. I suggest that most of the browsers don't check it anyway, otherwise they would not display contents with broken checksums.
Scripts seem to be a little bit faster, you may weigh up advantages with disadvantages (e.g. when you use a secure connection).
postmaster at ak-dev dot com ¶
20 years ago
It seems that you don't need anymore to add the crc and the size at the end of the stream, it works fairly with php4.06! I don't have tried for the moment with highest version than php 4.06.
The source code:
<?
function checkCanGzip() {
global $HTTP_ACCEPT_ENCODING;
if (headers_sent()) return 0;
if (strpos($HTTP_ACCEPT_ENCODING, 'x-gzip') !== false) return "x-gzip";
if (strpos($HTTP_ACCEPT_ENCODING,'gzip') !== false) return "gzip";
return 0;
}
function gzDocOut() {
if ($encoding = checkCanGzip()) {
$contents = ob_get_contents();
ob_end_clean();
header("Content-Encoding: ".$encoding);
print("\x1f\x8b\x08\x00\x00\x00\x00\x00");
$size = strlen($contents);
$contents = gzcompress($contents, 9);
$contents = substr($contents, 0, $size);
print($contents);
// print(pack('V', crc32($contents)));
// print(pack('V', $size));
exit();
}
else {
ob_end_flush();
exit();
}
}
ob_start();
ob_implicit_flush(0);
print("your stuff...");
gzDocOut();
?>
For the moment, I have only tried on IE5.5, and it works well.
Effectively, with the crc and the size at the end of gile, it doesn't work!
IT
PHP
编程语言
Linux
开发编程
科技
Elasticsearch
HTML/CSS/XML
网络
JAVA
NoSQL
数据库
面试
C/C++
Golang
Git
算法
操作系统
正则表达式
Redis
互联网
MySql
JavaScript
运维
软件
国际
架构设计
Mac OS
TCP/IP
Excel
Vim
Windows
Socket
Oracle
VR
MongoDB
Python
运营
MemCache
商业
硬件
电子
娱乐
设计
nginx
摄影
游戏
WordPress
HTTP
团建
数码电器
广告
--
php7 安装fileinfo扩展
preg_split — 通过一个正则表达式分隔字符串
php位值,解决 PHP 中 usort 在值相同时改变原始位置的问题
[PHP] inet_pton/inet_ntop IP地址转换函数
array_multisort() 多字段排序
PHPMailer设置utf8 PHPMailer字符集CharSet
ADORecordSet对象
adodb手册
PHP 8.1
[鸟哥]PHP_INT_MIN 和 -9223372036854775808
opcache预加载
Composer的Packagist资源
mysql面试题
[鸟哥]PHP FFI详解 - 一种全新的PHP扩展方式
PHP7添加redis扩展
PHP mkdir()写出来的权限与mode值不符合
Packagist / Composer 中国全量镜像
PHP获取指定函数定义在哪个文件中及行号
adodb连接mysql多个数据库的问题
常用的php ADODB使用方法集锦