MySQL COMPRESS vs PHP gzcompress - php

Question

I am developing a PHP application where large amounts of text needs to be stored in a MySQL database. Have come across PHP's gzcompress and MySQL's COMPRESS functions as possible ways of reducing the stored data size.
What is the difference, if any, between these two functions?
(My current thoughts are gzcompress seems more flexible in that it allows the compression level to be specified, whereas COMPRESS may be a bit simpler to implement and better decoupling? Performance is also a big consideration.)

Answer 1

The two methods are more or less the same thing, in fact you can mix them: compress in php and uncompress in MySQL and vice versa.
To compress in MySQL:
INSERT INTO table (data) VALUE(COMPRESS(data));
To compress in PHP:
$compressed_data = "\x1f\x8b\x08\x00".gzcompress($uncompressed_data);
To uncompress in MySQL:
SELECT UNCOMPRESS(data) FROM table;
To uncompress in PHP:
$uncompressed_data = gzuncompress(substr($compressed_data, 4));
Another option is to use MySQL table compression.
It only require configuration and then it is transparent.

Answer 2

This may be an old question, but it's important as a Google search destination. The results of MySQL's COMPRESS() vs PHP's gzcompress() are the same EXCEPT for MySQL puts a 4-byte header on the data, which indicates the uncompressed data length. You can easily ignore the first 4 bytes from MySQL's COMPRESS() and feed it to gzuncompress() and it will work, but you cannot take the results of PHP's gzcompress() and use MySQL's UNCOMPRESS() on it, unless you take specific care to add in that 4-byte length header, which of course requires having the uncompressed data already...

Answer 3

The accepted answer does not use the right 4 byte header.
The first 4 bytes are the LENGTH and not a static header.
I have no idea about the implications of using a wrong length but it can not be good and has the potential to crash the database or table content in the future (if not now)
The correct answer with POC example:
Output of mysql:
mysql : "select hex(compress('1234512345'))"
0A000000789C3334323631350411000AEB01FF
The php equivalent:

Answer 4

They both use zlib, so the compression will likely be about the same. Test it and see.

Answer 5

Adding this answer for reference, as I needed to use uncompress() to decompress data where the decompressed size was stored in a separate column to the blob.
As per the previous answers, uncompress() expects the first 4 bytes of the compressed data to be the length, stored in little-endian format. This can be prepended using concat e.g.
select uncompress(
concat(
char(size & 0x000000ff),
char((size & 0x0000ff00) >> 8),
char((size & 0x00ff0000) >> 16),
char((size & 0xff000000) >> 24),
compressed_data)) as decompressed
from my_blobs;

Answer 6

Johns answer is almost correct. The length must be computed by using strlen instead of mb_strlen as the latter will recognize multibyte characters as "1 character" although they span multiple bytes. Take the following example with a "▄" character that consists of 3 bytes:
$string="▄";
$compressed = gzcompress($string, 6);
echo "with strlen\n";
$len = strlen($string);
$head = pack('V', $len);
$base64 = base64_encode($head.$compressed);
echo "Length of string: $len\n";
echo $base64."\n";
echo `mysql -e "SELECT UNCOMPRESS(FROM_BASE64('$base64'))" -u root -proot -h mysql`;
echo "\n\nwith mb_strlen\n";
$len = mb_strlen($string);
$head = pack('V', $len);
$base64 = base64_encode($head.$compressed);
echo "Length of string: $len\n";
echo $base64."\n";
echo `mysql -e "SELECT UNCOMPRESS(FROM_BASE64('$base64'))" -u root -proot -h mysql`;
Output:
with strlen
Length of string: 3
AwAAAHicezStBQAEWQH9
UNCOMPRESS(FROM_BASE64('AwAAAHicezStBQAEWQH9'))
▄
with mb_strlen
Length of string: 1
AQAAAHicezStBQAEWQH9
UNCOMPRESS(FROM_BASE64('AQAAAHicezStBQAEWQH9'))
NULL