Difference between character and string when constructing a 1-d array of the specified type

user_6681701 0
Viewed 61

I am confusing with when constructing a 1-d array of the specified type by usung getindex(type[, elements...]).

Of course, that I can convert Int 8 when the element are Int

getindex(Int8, 1, 2)

2-element Vector{Int8}:
1
2

Even when the element are character format, I can convert it to Int8 :

getindex(Int8, '1', '2')

2-element Vector{Int8}:
49
50

However, I can not convert when the element are in string format.

getindex(Int8, "1", "2")

and, raise the following error :

MethodError: Cannot `convert` an object of type String to an object of type Int8
Closest candidates are:
  convert(::Type{T}, ::Ptr) where T<:Integer at pointer.jl:23
  convert(::Type{IT}, ::GeometryBasics.OffsetInteger) where IT<:Integer at C:\Users\Admin\.julia\packages\GeometryBasics\WMp6v\src\offsetintegers.jl:40
  convert(::Type{T}, ::SentinelArrays.ChainedVectorIndex) where T<:Union{Signed, Unsigned} at C:\Users\CARVI\.julia\packages\SentinelArrays\tV9lH\src\chainedvector.jl:209
  ...

Stacktrace:
 [1] setindex!(A::Vector{Int8}, x::String, i1::Int64)
   @ Base .\array.jl:839
 [2] getindex(#unused#::Type{Int8}, x::String, y::String)
   @ Base .\array.jl:393
 [3] top-level scope
   @ In[35]:1
 [4] eval
   @ .\boot.jl:360 [inlined]
 [5] include_string(mapexpr::typeof(REPL.softscope), mod::Module, code::String, filename::String)
   @ Base .\loading.jl:1116

Why getindex() allow character element to convert to different format (like character -> Int8), but string ?

julia
1 Answers

First of all, that's a rather weird way of writing array literals: getindex(T, xs...) is usually written as T[xs...].

However, the error already quite clearly tells you what went wrong:

Cannot convert an object of type String to an object of type Int8

How do you imagine a general conversion from String to Int8 to look like? What 8-bit integer should correspond to the string "slkdfjls", for example? A string is after all a pretty much arbitrary sequence of bytes. And contrary to your expectation, Julia does not make an attempt to do any parsing of the contained value (for that, use parse(Int8, "1").

Characters on the other hand represent (if valid) single UTF-8 code points, and it is meaningful to reinterpret their fixed amount of bytes as a number:

julia> convert(Int16, '†')
8224

julia> convert(Int8, '1')  # certainly not Int8(1)!
49

The conversion is already borderline meaningful when the value exceeds the range of the target type:

julia> convert(Int8, '†')
ERROR: InexactError: trunc(Int8, 8224)
...

UTF-8 characters that happen to be representable by only one byte can be losslessly converted to Int8; this covers all of ASCII. Above that, an error is raised. This is no different from convert(Int8, Int32(something)).

user_13462760
Related