TAB-completion inside a string (inside quotes) is different from normal

[KronosTheLate]

When writing a string, e.g. with the cursor where the star is in md"my_text* ", the autocompletion is different compared to when not in a string. When trying to autocomplete \omega to get the greek symbol, it is autocompleted to \Documents and Settings\\.

This is annoying in general, but especially when defining a unit like Ohm. I then have to close the string macro with an extra ", write \Omega<TAB>, and erase the extra ". It is not a major problem at all, but annoying. Anyone else experiencing this, and does anyone know of a way to fix it?

[fonsp]

Thank you for the report!

This is supposed to be a 'feature' of Julia: autocompleting paths inside strings, but I'm not sure that I agree with this feature. Especially because (in the future) I want to motivate authors to write notebooks that are standalone, i.e. do not depend on local files.

@aviatesk What do you think? Can/Should we disable path autocomplete inside strings?

FYI it's designed for this situation:

my_text = read("em<TAB>

will autocomplete to

my_text = read("emma.txt

[aviatesk]

Can/Should we disable path autocomplete inside strings?

you can filter our path completions:

Pluto.jl/src/runner/PlutoRunner.jl

Lines 684 to 690 in 9bea6ea

	results, loc, found = completions(query, pos, workspace)
	if endswith(query, '.')
	# we are autocompleting a module, and we want to see its fields alphabetically
	sort!(results; by=(r -> completion_text(r)))
	else
	filter!(≥(0) ∘ my_score, results) # too many candiates otherwise
	end

with something like

filter!(!Base.Fix2(isa, FuzzyCompletions.PathCompletion), result)

And query parameter there allows further fine tunings like filtering path completions only within string-macro scope, etc:

# suppress path completions only within string-macro scope
x = Meta.parse(query)
if isa(x, Expr) && x.head === :macrocall
     filter!(!Base.Fix2(isa, FuzzyCompletions.PathCompletion), result)
end

Well, I'm not sure if we should disable path completions or not. Even within a notebook context, it's sometimes useful imho e.g. when we want to load local CSV data when using Pluto in a data science context.

[aviatesk]

I then have to close the string macro with an extra ", write \Omega, and erase the extra "

Well, separate from the path completion discussion, FuzzyCompletions somehow doesn't seem to provide latex completions within a string scope. I will dug into this this weekend or so.

[KronosTheLate]

I then have to close the string macro with an extra ", write \Omega, and erase the extra "

Well, separate from the path completion discussion, FuzzyCompletions somehow doesn't seem to provide latex completions within a string scope. I will dug into this this weekend or so.

Amazing, thanks a lot <3. This package and team are amazing - I have recently swiched to Pluto as my main IDE/place to do programming ^_^

[aviatesk]

FuzzyCompletion@0.4.0 will fix this issue.

[fonsp]

Can you make a PR? Is there a breaking change?

[aviatesk]

Nothing breaking, but here is: #727

[fonsp]

Thanks! Next time, you could register it as a patch (0.4.0 -> 0.4.1) instead of a breaking release (0.4.0 -> 0.5.0). This means that Atom and Pluto get the latest FuzzyCompletion features without doing anything :)

[aviatesk]

I hesitated to do patch release and silently fix this issue , but yeah, it certainly doesn't follow the semantic versioning..

[icweaver]

Hi, sorry if this is the wrong place to ask this. Has the file completion for paths already started to be removed? I think I am still able to complete paths if I type a character that is present:

"em<TAB>

completes to:

"emma.txt

but attempting to complete from nothing just inserts the tab:

"<TAB>

completes to:

"    |<-- cursor now here

I totally see your point about trying to nudge people towards writing standalone notebooks, but I personally happen to fall in the data science context brought up by @aviatesk:

Well, I'm not sure if we should disable path completions or not. Even within a notebook context, it's sometimes useful imho e.g. when we want to load local CSV data when using Pluto in a data science context.

and would absolutely love to retain this feature if it's not too late

[fonsp]

@icweaver Does "./<TAB>" still work? Then we'll leave it at that

[icweaver]

Hey Fons, thanks for getting back to me so quickly! Unfortunately not. It seems to just insert a single space instead

[icweaver]

Oh, but doing it without the forward slash does the trick! I'm happy to just stick with that =]