Explain Perl code

What this Perl sample shows

scalars, arrays and hashes

The first character says what a variable holds: $line is one value, @slowest a list and %hits a hash of keys and values. One entry of a hash is a single value, so it takes $: $hits{$path}. The first time a path turns up its entry is empty, and ++ starts it from 0.

regular expressions

m{...} is a match written with braces in place of slashes, which helps when the text has / in it. Each pair of brackets captures one part into $1 to $4: the method, the path, the 3-digit status and the milliseconds. next unless skips a line that doesn't match, so the sample log's health check ok line is ignored.

strict and warnings

use strict makes every variable be declared with my. Without it, a typo such as $hit{...} for $hits{...} quietly made a second hash and the count came out as 0 paths. With it, Perl refused to run and named %hit.

sort with a comparison

sort { ... } keys %hits orders the paths using the block, where $a and $b are the two being compared. <=> compares numbers, and putting $b on the left sorts largest first, here by average time. @slowest[0 .. 2] takes the first three, and last unless defined $path stops early when the log has fewer pages.

Good to know

• Run with Perl 5.42.2 on a 9-line sample log, the script printed /api/signup at 925 ms average over 2 requests, /pricing at 396 ms and /docs at 92 ms, then 1 server errors for the one 503.
• <> reads the files named on the command line a line at a time, or standard input when there are none, so cat access.log | perl slow_pages.pl printed the same result.
• With only two pages in the log, the script printed both and stopped with no warning, because of the defined check.

Watch out: == compares numbers and eq compares text. "GET" == "POST" came out true: Perl turned both words into 0, printed Argument "POST" isn't numeric in numeric eq (==) and carried on. A log saved on Windows has a hidden \r at the end of each line, so ms$ never matched and the script printed only 0 server errors, with no error. Adding $line =~ s/\r$//; as the loop's first line fixed it.