Question

[regex]::replace('test test','^(.*?)test', 'barf')

returns ‘barf test’

Why doesn’t it replace all occurrences of ‘test’? This must have something to do with the position at which a subsequent replace iteration begins.

Answer 1

Quick answer: you anchored it at the beginning of the input (^) and your first group ((.*?)) did not capture anything (since the first occurrence of test was found right after the beginning of line and you use a lazy quantifier — furthermore you don’t use the capture in your replacement string. Had you used a “normal” quantifier, the last occurrence of test would have been replaced).

Long answer: a regex never needs to match the whole input, only the parts which are necessary. What’s more, when cycling through an input, the regex engine will start the next round from the position where it successfully completed a match.

Here, you want to replace a sequence of characters which is test. Note that it will also means that testosterone will be matched (or untested). If you want to match test as a “word”, use the word anchor \b.

This works (tested on Powershell v2):

[regex]::replace('test test','\btest\b', 'barf')

The engine in action looks something like this:

# beginning
regex: |\btest\b
input: |test test
# \b: matched,  beginning of input followed by word character
regex: \b|test\b
input: |test test
# literal matching of t, then e, then s, then t
regex: \btest|\b
input: test| test
# \b: match, word character followed by non word character
regex: \btest\b|
input: test| test
# replacement
regex: \btest\b|
input: barf| test
# beginning of second round
regex: |\btest\b
input: barf| test
# \b: match, word character followed by non word character
regex: \b|test\b
input: barf| test
# t: not matched. Failed matching. Proceeding to next character
regex: |\btest\b
input: barf |test
# \b: match
regex: \b|test\b
input: barf |test
# literal matching of t, then e, then s, then t
regex: \btest|\b
input: barf test|
# \b: match, word character followed by end of input
regex: \btest\b|
input: barf test|
# replacement
regex: \btest\b|
input: barf barf|
# beginning of next round
regex: |\btest\b
input: barf barf|
# end of input: end of processing