Fix print_script for 1.10

[j-fu]

Appearantly, at the time when @generated is evaluated, methods from other packages may not have been compiled yet, so
hasmethod(show, Tuple{IO, MIME{Symbol("text/javascript")}, value}) returns false, and the @generated function print_script throws an error.

As the old version worked in 1.9.3, the different behavior appears to be the result of subtle changes in precompilation/initialization behavior somewhere between 1.9.3 and 1.10.0-beta2.

[j-fu]

This should fix #35 and at once JuliaPluto/AbstractPlutoDingetjes.jl#12 , JuliaLang/julia#51476 and j-fu/PlutoVista.jl#13 .

[j-fu]

Working MWE for this PR

[j-fu]

The change in behavior appears to be due to the new parser.

When setting export JULIA_USE_FLISP_PARSER=1 the original MWE using the released version of HypertextLiteral works.

[clarkevans]

I'm curious does the new parser fix JuliaLang/julia#18221 ? If so we could report it as being fixed.

Regardless, could wrap this in a version check? This code is there to reduce user confusion.

[j-fu]

No it doesn't solve JuliaLang/julia#18221 .
As for the generated function based workaround, julia docs state that generated functions should not use hasmethod. So the new parser seems to be in accord with the docs here.

Will see if this can be wrapped into a version check. But in 1.10 we are on field 1 anyway.

[j-fu]

What about implementing

Base.show(io::IO, ::MIME"text/javascript", value::Any)=error("""Missing show(::IO,::MIME"text/javascript",::$(typeof(value))""")

Seems to be a bit nasty as well, but might work (did not test this yet outside of the REPL).

[clarkevans]

That would be type thievery?

[j-fu]

I think not exactly, but kind of scary anyway.

I added the version discriminator as you proposed. I think this is ready to merge.

[clarkevans]

I suppose there's not much else we could do.