Define islowercase for String

[jariji]

In 1.7 islowercase is defined for characters but not strings. It would be useful to have for strings.

(From discourse.)

[PallHaraldsson]

FYI:

julia> all(islowercase, "Páll")
false

julia> all(islowercase, "páll")
true

# first I thought it might be, but it (and the above) not obvious to all:

julia> islowercase.("Páll")
ERROR: MethodError: no method matching islowercase(::String)

# since non-obvious feel free to make a PR for, and corresponding isuppercase:

julia> import Unicode.islowercase

julia> islowercase(s::AbstractString) = all(islowercase, s)

[KristofferC]

I also think lowercase has to do with a character and not a string. Using all together with islowercase is therefore the natural way of expressing this.

[oscardssmith]

IMO, it's very odd that lowercase("Hi") works but islower("Hi") doesn't. Adding to triage to decide what we want to do about this.

[stevengj]

On the other hand, the definition that people seem to want in many cases is:

all(c -> lowercase(c) == c, s) # equivalent to lowercase(s) == s

which is not the same thing as all(islowercase, s):

This ambiguity may be a good argument for not defining islowercase(::String).

[stevengj]

Consider

julia> "élan" == lowercase("élan")
true

julia> all(c -> c == lowercase(c), "élan")
true

julia> all(islowercase, "élan")
false

If you are confused by this, realize that "élan" includes the unicode combining character U+0301. On the other hand, if you use the NFC normalization "élan" of this string (which uses the accented character U+00E9), you get:

julia> all(islowercase, "élan")
true

(But if you define islowercase(s::String) = s == lowercase(s), then it will be inconsistent with the definition of islowercase for characters.)

[gbaraldi]

julia> all(islowercase, "élan")
false

Is probably a wrong result for most people. which is something to worry about

[stevengj]

@gbaraldi, is this the "wrong" result:

julia> all(islowercase, "1")
false

julia> islowercase('1')
false

julia> all(islowercase, "́") # "\acute"
false

julia> islowercase('́') # '\acute'
false

?

[JeffBezanson]

We explicitly decided to remove these kinds of predicates for strings in v0.6.

[PallHaraldsson]

Does it make sense to rather define as (and vise versa):

islowercase(s::AbstractString) = any(!isuppercase, s)  # also faster, and different

[stevengj]

We explicitly decided to remove these kinds of predicates for strings in v0.6.

In particular, this was done in #20342, following the discussion in #14156.

Does it make sense to rather define as (and vise versa): islowercase(s::AbstractString) = any(!isuppercase, s)

No, because then islowercase("1") will return true and islowercase('1') will return false, which seems inconsistent.

In general, the fact that there are so many variations on what it might mean for a string to "be lowercase" (e.g. all(islowercase, s) or !any(isuppercase, s) or any(!isuppercase, s)) is a good argument by itself for not defining this method.

[stevengj]

Note that Julia isn't alone here. See e.g. Swift’s string.lowercased() method and char.isLowercased property. Or the C# String.ToLower() and Char.IsLower methods. Or the Go ToLower(str) and IsLower(char) functions. Or …

The odd one out is Python, which does have a string.islower() method (equivalent to all(islowercase, string)), but that's probably because Python doesn't have a character type.

[JeffBezanson]

I think this can be closed. I don't see an overwhelmingly convincing reason to revisit this decision.