RFC: move stringmime to Base64, rename reprmime -> repr

[stevengj]

This PR makes three main changes:

• It renames reprmime(mime, x) to repr(mime, x), analogous to how we merged writemime and show.

• It moves stringmime to the Base64 package, since the only reason to use this function rather than repr is if you want base64 encoding for binary MIME types.

• It adds optional context::Pair{Symbol}... arguments context keywords to repr to repr(mime, x) and stringmime, similar to c418a10 and sprint.

One slight wrinkle is that there is an ambiguity for repr(mime::Union{MIME,AbstractString}, p::Pair{Symbol}): are you asking for repr(mime) using the context p, or repr(p) using the MIME type mime? I opted to resolve this as repr(p) for mime::MIME and repr(mime) for mime::AbstractString.

To do:

• Tests of the new context arguments
• NEWS

[stevengj]

I am having a bootstrapping problem with moving stringmime to Base64, which means that the Markdown package now imports Base64:

LoadError("sysimg.jl", 514, LoadError("/Users/stevenj/Documents/Code/julia/usr/share/julia/site/v0.7/Test/src/Test.jl", 32, LoadError("/Users/stevenj/Documents/Code/julia/usr/share/julia/site/v0.7/InteractiveUtils/src/InteractiveUtils.jl", 16, LoadError("/Users/stevenj/Documents/Code/julia/usr/share/julia/site/v0.7/Markdown/src/Markdown.jl", 9, ArgumentError("Module Base64 not found in current path.\nRun `Pkg.add(\"Base64\")` to install the Base64 package.")))))

I don't understand why stdlib/Test can do using InteractiveUtils but stdlib/Markdown can't do using Base64. What am I missing?

[JeffBezanson]

You need to add an entry for Base64 to Markdown's Project.toml file.

[stevengj]

It looks like most of the calls to stringmime are for text/plain anyway, which means they can be replaced with repr calls.

[StefanKarpinski]

The error messages for code loading are going to need a bit of work before we cut a final release. And documenting how all this stuff works will help, of course.

[nalimilan]

😆

[stevengj]

32-bit CI failures seem to be unrelated network problems.

[StefanKarpinski]

I've started opening issues to track the known ongoing CI failures: ci Continuous integration . Shocking idea, I know.

[JeffBezanson]

I think we should resolve the ambiguity in repr by adding a context keyword argument (like sprint has).

[JeffBezanson]

(The Pair arguments to repr were added very recently.)

[StefanKarpinski]

You can replace repr(obj, pairs...) with repr(obj; context = ...) the way that sprint takes; alternatively, we could just replace the one usage of this with sprint(show, x, context = (:module => nothing)). There's one other usage of this but it's in Pkg3 and I'll know what to do when it breaks, so don't bother with a deprecation.

[stevengj]

Updated to a context keyword. I also added keyword documentation to sprint, and added a context keyword to base64encode.

[StefanKarpinski]

This line's also going to have to change:

julia/base/loading.jl

Line 1119 in bde667d

append!(Base.LOAD_PATH, $(repr(LOAD_PATH, :module => nothing)))

The simplest fix is changing it to

repr(LOAD_PATH, context = (:module => nothing))

[stevengj]

Looks like the usual unrelated CI failures.