Testing about PHP `stream_get_line` and `fgets` for large files

stream_get_line.md

Testing about PHP stream_get_line and fgets for large files

From http://ie2.php.net/manual/en/function.fgets.php#113113

Regarding Leigh Purdie's comment (from 4 years ago) about stream_get_line being better for large files, I decided to test this in case it was optimized since then and I found out that Leigh's comment is just completely incorrect fgets actually has a small amount of better performance, but the test Leigh did was not set up to produce good results.

The suggested test was:

$ time yes "This is a test line" | head -1000000 | php -r '$fp=fopen("php://stdin","r"); while($line=stream_get_line($fp,65535,"\n")) { 1; } fclose($fp);'

0m1.616s

$ time yes "This is a test line" | head -1000000 | php -r '$fp=fopen("php://stdin","r"); while($line=fgets($fp,65535)) { 1; } fclose($fp);'

0m7.392s

The reason this is invalid is because the buffer size of 65535 is completely unnecessary piping the output of "yes 'this is a test line'" in to PHP makes each line 19 characters plus the delimiter so while I don't know why stream_get_line performs better with an oversize buffer, if both buffer sizes are correct, or default, they have a negligable performance difference - although notably, stream_get_line is consistent - however if you're thinking of switching, make sure to be aware of the difference between the two functions, that stream_get_line does NOT append the delimiter, and fgets DOES append the delimiter.

Here are the results on one of my servers:

Buffer size 65535
stream_get_line:    0.340s
fgets:   2.392s

Buffer size of 1024
stream_get_line:  0m0.348s
fgets: 0.404s

Buffer size of 8192 (the default for both)
stream_get_line: 0.348s
fgets:  0.552s

Buffer size of 100:
stream_get_line: 0.332s
fgets: 0.368s

Source http://ie2.php.net/manual/en/function.fgets.php#113113