add chopprefix, chopsuffix

[kcajf]

Attempt to address #38477

[stevengj]

Maybe chophead and choptail?

[stevengj]

This may not be correct if s and prefix are different subtypes of AbstractString, since in that case lastindex(prefix) may not be a correct index into s.

[stevengj]

Probably we need two methods: one for AbstractString arguments, and an optimized method for Union{String,SubString{String}} arguments. (In the latter case you can just use 1+ncodeunits(prefix) as the starting index.)

[stevengj]

This is fine in principle, but is a little suboptimal because length(suffix) and prevind require two extra O(length(suffix)) loops. In the case where both s and suffix are of type String or SubString{String}, one option is to use thisind(s, ncodeunits(s) - ncodeunits(suffix)).

(As above, we probably want two methods here, one optimized for String.)

[JeffreySarnoff]

+1 for chophead, choptail
or removeprefix, removesuffix

[kcajf]

Yes, I like chophead/choptail. Thanks @stevengj for the feedback, I'll add those optimisations shortly.

[thchr]

It might be nice to avoid introducing a new function name entirely and instead just add this as a new method for chop or lstrip/rstrip (that would feel more in tune with multiple dispatch to me)?
Right now, there's already a fair number of different options for changing the ends of strings, whose names nevertheless invoke pretty much the same idea in my mind:

• chop: remove n characters from tail/head.
• chomp: remove \n from tail of string.
• lstrip/rstrip: remove specific character(s) or predicate from tail/head of string.

I always have trouble remembering which of these applies to any given situation. It would be nice if this could be added without adding more similarly sounding function names for slightly different versions of "get rid of something at the ends of a string".

E.g., could adding this to lstrip/rstrip work, i.e. as a ::AbstractString, ::AbstractString method? Then it could potentially also be generalized later to stripping multiple prefixes/suffixes sequentially as a ::AbstractString, Vector{<:AbstractString} method.

[StefanKarpinski]

The names chophead and choptail sound to me like they unconditionally remove something, which matches the behavior of chop(str, head=3) and chop(str, tail=3). That does suggest that maybe we could do

chop(str, head="prefix")
chop(str, tail="suffix")

Although those suggest to me that they might error if str doesn't begin with prefix or end with suffix.

The trouble with defining lstrip(::AbstractString, ::AbstractString) is that we already have this:

julia> rstrip("Hello", collect("lo"))
"He"

I actually expected rstrip("Hello", "lo") to work and do the same thing. It doesn't, but I think that since we treat strings as collections of characters, it would be bad to give that a different meaning.

Overall my favorite so far is chop(str, head="prefix") and chop(str, tail="suffix").

[vtjnash]

Stylistically, I think I much prefer chophead and choptail to kwargs for that. Using named arguments instead of positional for APIs like this (where the order can be confused) looks helpful, but seems inconsistent to change now.

[vtjnash]

I think we can do this fairly compactly as follows (startswith could also be written this way, to show they are the same):

Suggested change

	if startswith(s, prefix)
	SubString(s, nextind(s, lastindex(prefix), 1), lastindex(s))
	else
	SubString(s)
	end
	i, j = iterate(a), iterate(b)
	while true
	j === nothing && i === nothing && return SubString(s, 1, 0) # s == prefix: empty result
	j === nothing && return SubString(s, i[2]) # ran out of prefix: success!
	i === nothing && return SubString(s) # ran out of source: failure
	i[1] == j[1] || return SubString(s) # mismatch: failure
	i, j = iterate(a, i[2]), iterate(b, j[2])
	end

Then the general chopend is pretty much the same, but with slight changes (such as calling iterate(Iterators.Reverse(a)) at the top, and SubString(s, firstindex(s), i[2]) on success)

[vtjnash]

I decided to just go ahead and take this over and finish it. I think it is ready to merge, but can add to triage if (@StefanKarpinski) we want to discuss merging chop|chopprefix|chopsuffix into chop(s; [head], [tail] [prefix], [suffix]) and deciding which order to apply those in.

[kcajf]

Thanks @vtjnash!

[stevengj]

LGTM, modulo bikeshedding the names.

[oscardssmith]

Lets bikeshed the names.

[JeffBezanson]

Triage says 👍

[stevengj]

Was this PR missing a NEWS?