Faster dump by less tiny writes()

[staabm]

while writing the dump the script is doing millions of tiny writes().

after this PR we are buffering the contents within a string a do the write only after we reached the net-buffer-length. this means we are doing waaaay less IO calls.

before this PR:

mstaab@mst22:/cluster/www/www/www/mysqldump-php$ time php test2.php

real    0m14.194s
user    0m10.762s
sys     0m2.340s

after this PR

mstaab@mst22:/cluster/www/www/www/mysqldump-php$ time php test2.php

real    0m7.492s
user    0m6.403s
sys     0m0.204s

which is now 40-50% faster 🚀

[stempora]

What if you have very large tables? In my case i have a table of ~90gb and growing

[staabm]

Give it a try and report back :-)

[ifsnop]

Hi, thanks for the PR and the time you took to evaluate this. During your tests... did you use compression output or was it disabled? Please see #284.

[staabm]

As far as I remember I used default settings / not sure whether this means compressed or not

[ifsnop]

By default, compression is configured to None. So I think the option to buffer only when compression is None is better, let me do some tests and I will update the PR and fix the issue.

Thanks for your effort.

Message ID: @.***>

[ifsnop]

This makes more sense than #284.