`isvalid` error with PythonCall.Utils.StaticString

[CarloLucibello]

When converting to PyArray some numpy arrays with string elements, I end up with strings of type PythonCall.Utils.StaticString{UInt32, 44}.

It seems that this type is not well supported though, and I get the following errors when indexing and computing the length:

julia> s = PythonCall.Utils.StaticString{UInt32, 44}("ababababb")
"ababababb"

julia> s[1]
ERROR: MethodError: no method matching isvalid(::PythonCall.Utils.StaticString{UInt32, 44}, ::Int64)
Closest candidates are:
  isvalid(::AbstractString, ::Integer) at strings/basic.jl:143
  isvalid(::SubstitutionString, ::Integer) at regex.jl:566
  isvalid(::Type{T}, ::Integer) where T<:AbstractChar at strings/unicode.jl:58
  ...
Stacktrace:
 [1] isvalid
   @ ./strings/basic.jl:143 [inlined]
 [2] getindex(s::PythonCall.Utils.StaticString{UInt32, 44}, i::Int64)
   @ Base ./strings/basic.jl:186
 [3] top-level scope
   @ REPL[5]:1

julia> length(s)
ERROR: MethodError: no method matching isvalid(::PythonCall.Utils.StaticString{UInt32, 44}, ::Int64)
Closest candidates are:
  isvalid(::AbstractString, ::Integer) at strings/basic.jl:143
  isvalid(::SubstitutionString, ::Integer) at regex.jl:566
  isvalid(::Type{T}, ::Integer) where T<:AbstractChar at strings/unicode.jl:58
  ...
Stacktrace:
 [1] isvalid
   @ ./strings/basic.jl:143 [inlined]
 [2] length
   @ ./strings/basic.jl:396 [inlined]
 [3] length(s::PythonCall.Utils.StaticString{UInt32, 44})
   @ Base ./strings/basic.jl:387
 [4] top-level scope
   @ REPL[6]:1

julia> s isa AbstractString # in principle the type should implement the whole interface
true

[cjdoris]

Yeah, PythonCall.Utils.StaticString is pretty minimal in what it supports - you can convert them to String but not much else.

Supporting indexing is harder because it requires understanding UTF-8 and UTF-32 - AFAIK Julia doesn't provide much support for implementing your own string types in terms of handling the encoding (even UTF-8) except for the transcode function, which I use to convert to String.

If you are getting these from a PyArray, I have just pushed a change to the main branch so that you can do PyArray{String}(x) on a Python array of strings, which converts the elements to String when you access them. Of course this is slower than accessing them as PythonCall.Util.StaticString but if you need indexing then it's an option.

[CarloLucibello]

I see you have implemented indexing in the last commits, thanks, the examples in the OP now work:

julia> s = PythonCall.Utils.StaticString{UInt32, 44}("ababababb")
"ababababb"

julia> s[1]
'a': ASCII/Unicode U+0061 (category Ll: Letter, lowercase)

julia> length(s)
9

julia> s[1:2]
"ab"

julia> s[1:2] |> typeof
SubString{PythonCall.Utils.StaticString{UInt32, 44}}

Can we get a new release?

[cjdoris]

I decided it wasn't so hard! Sure I'll make a release this week.

[CarloLucibello]

Hi, sorry to bother but it would be great to have a new release!

[cjdoris]

Sure - done.