Tunde Jacobs

I start every explanation with the sample line, printed in full, because a pattern written without the input is a guess. The construction that matters is the non-greedy quantifier. A pattern matching up to the next delimiter will, without it, run to the end of the line and return the whole remainder. Non-greedy stops at the first match, which is usually what was meant, and the difference between the two is invisible until the output is long.

Excluding a character class is how you match a field that contains no delimiter. That is more precise than matching the negated delimiter and faster, and it fails loudly when the assumption is wrong rather than quietly swallowing a field. Groups are how you capture parts of a match, and they are also how a pattern gets slow.

Every group adds work, so an unnecessary capturing group inside a repeated element is a measurable cost. Unnamed groups still capture unless marked otherwise, so a group added for readability shifts every numbered reference after it. Failure is worth planning for.

A pattern that does not match returns nothing, which looks identical to a line that does not contain the field. Counting matches per line rather than total matches catches that. I finish on the volume limit, because a nested quantifier over a large file can stall, and every pattern I publish has been run against something larger than the example. Every pattern on this page has been run against a file larger than the example, and I say roughly how large, because a pattern that works on ten lines is not yet tested.