doctests for the manual, part 3

[fredrikekre]

nothing exciting here, just plain conversion to doctests and some whitespace fixes

(make check-whitespace and the doctests pass locally, remove [ci skip] from commit message if/when this is merged)

[fredrikekre]

this was deprecated in #19905

[tkelman]

is the replacement syntax demo'ed in the docs anywhere?

[fredrikekre]

Not that I know of.

[tkelman]

should it be?

[fredrikekre]

It seems that it is only when paired with assignment that is deprecated:

julia> global a, b

julia> global x = 1; d

WARNING: deprecated syntax "global c=1, d" at REPL[2]:1.
Use "global c=1; global d" instead.

It seems a bit overkill to say that if you wanna pair it with assignment the keyword needs to be in front of each variable?

On another note: There is an empty line before the warning which is not the case for other warnings

[tkelman]

The extra newline should probably be fixed... syntax deprecations get thrown from inside the parser I think. In 1.0 I guess this syntax would be an error, in which case hopefully the error message will indicate the correct syntax in a clear way?

[tkelman]

why are you removing backtrace lines?

[fredrikekre]

The point of the example is just to show that it throws a DomainError? And if the backtrace is removed like this the doctest doesn't need to be updated every time something change in math.jl
Should I include backtraces for all examples then? I have omitted it for some other blocks too.

[tkelman]

I find at least the first line or two of the backtrace to be pretty useful, it tells you where to go find how the code is implemented if you want to look, and it's representative of what happens when you run this code yourself. The line numbers thing we should address in terms of how we run the doctests and how picky we are about everything matching. I think we should detect line numbers in the output and be permissive of them changing, everywhere except on release-branch builds.

[KristofferC]

As a counter point I typically find all the backtraces in the docs annoying and noisy.

[fredrikekre]

It can be noted that the [...] is actually rendered too so it is implicit that there are things hidden. The above case renders to this:

Note also that it is not consistent whether the backtrace is shown or not across the manual. What do other people have to say?

[MichaelHatherly]

What do other people have to say?

I've got no particular preference for either, largely depends on the purpose of each doctest I think.

[tkelman]

the space between the + and y should be kept

[MichaelHatherly]

Same remark, again, as in https://github.com/JuliaLang/julia/pull/20183/files#r97537182 can apply here as well I think.

[fredrikekre]

I guess no one else seem to be bothered by the stacktraces in the manual, so I will add them back.
But for the record I think that it is clear that the output is truncated from the dots, ref #20202 (comment)

I will update, and add them back.

[fredrikekre]

Re-added stacktraces, added doctests and check-whitespace pass locally.