REPL: allow tweaking the implicit IOContext of the REPL #29249

rfourquet · 2018-09-18T16:20:15Z

This adds an iocontext::Dict{Symbol,Any} field to LineEditREPL,
which can be initialized with atreplinit, and updated
interactively, e.g. Base.active_repl.iocontext[:compact] = true.

Fixes #20509.

An alternative would be providing specific functions, e.g. setiocontext(:key, value), resetiocontext(:key) or something. I think I prefer the dictionnary approach, because a Dict already has a convenient API for the needs here, and as an interactive feature, this can be changed relatively easily if needed.

I wonder whether we should think about a namespace policy to use this, as package X could say for example "use iocontext[:format] = :bin for printing with binary digits", while package Y could say have iocontext[:format] = whatever to control formatting of other stuff, and it becomes impossible to have control over both packages at the same time.
Another option would therefore to have a predetermined list of settable properties (:compact, :limit etc), in which case using a mutable struct with corresponding fields would probably be a more appropriate implementation than a Dict.

rfourquet · 2019-09-17T06:35:43Z

Almost one year later... I miss this functionality, for example I worked with small arrays yesterday which often don't fit the screen but whose content I want to inspect, I know of two ways to do that show(stdout, MIME("text/plain"), ans) and for x in ans; println(x); end, I end up doing the latter, but writing once Base.active_repl.iocontext[:limit] = false or Base..active_repl.iocontext[:displaysize] = (200, displaysize(stdout)[2]) (to increase the limit) would be great.

As an interactive feature, this does not bind us much, if we want to evolve the API later. Any objections?

rfourquet · 2019-09-25T12:07:45Z

Bump. I will be tempted to just merge if no-one expresses objections/suggestions.

KristofferC · 2019-09-25T12:29:18Z

In its current state it isn't really usable by anyone really (no docs) but perhaps you plan on adding that later?

rfourquet · 2019-09-25T15:13:42Z

Good point thanks, docs just added.

rfourquet · 2019-09-27T09:57:58Z

The failed test look unrelated, but it would be good if someone can confirm.
Also, someone willing to have a look, at least at the docs?

StefanKarpinski · 2019-10-01T15:18:36Z

Bump.

KristofferC · 2019-10-01T15:20:25Z

stdlib/REPL/docs/src/index.md

+In particular, the `:limit` attribute is set to `true`.
+Other attributes can receive in certain `show` methods a default value if it's not already set,
+like `:compact`.
+It's possible to specify the attributes used by the REPL via the `Base.active_repl.iocontext`


Is this the first time Base.active_repl is introduced? In that case, it might make sense to create a dedicated function for this in the REPL module.

It seems to be the first time that Base.active_repl is introduced. I'm not clear of the point of a dedicated function to access it, and I think it's fine allow it for now (this field is likely to stay for a while). That said, I should probably document how to do that in the startup file.

Well, the point is to not expose some weird global variable in Base that is only defined in at a certain point in the startup procedure.

Also, there is an options field in active_repl (unfortunately not yet documented), whose fields can be interactively explored with tab completion. AFAICT, if active_repl was accessible only via a call to a function, there would not be autocompletion.

If we put this in the documentation then the active_repl.iocontext is an official API and it can never ever be changed. It seems likely we want to have some better way to set options in the REPL than this in the future so committing to this "setting fields in a global variable in Base API" doesn't seem that great.

If there isn't a good way of setting options in the REPL right now I would prefer merging without exposing the current implementation as a stable API.

But I won't define the function get_active_repl() = Base.active_repl.

Heh, no.

My thinking is something along the lines of having:

REPL.options.iocontext = Dict(:compat => false) REPL.options.bell = false ...

where REPL.options is documented (and tab completable).

The active repl creates a reference to the options object when it gets initialized and the options object is always defined.

OK, this is beyond the scope of this PR though. Do you suggest that REPL.options be global? It could be an object with overloaded setproperty! and getproperty, to make it a proxy to Base.active_repl.options... Indeed looks like it may be better than Base.active_repl.

If we put this in the documentation then the active_repl.iocontext is an official API and it can never ever be changed.

OK, this was not my understanding, as it's interactive. So what about, in this PR, leaving the atreplinit portion of the doc as the official way, and pointing at the Base.active_repl as an experimental way to play with the options?

Sure, if we denote it as experimental, that's fine.

So I just marked the whole feature as experimental anyway, as once it's used, we may find out a better way to achieve the same, but we just need to start somewhere.

@KristofferC Just FYI Module.CONFIG is now used in at least three packages for this kind of purpose (disclaimer: I did the implementation in all cases):

https://github.com/queryverse/ElectronDisplay.jl
https://github.com/JuliaDebug/Cthulhu.jl
https://github.com/tkf/InteractiveCodeSearch.jl

Although I guess I'll follow stdlib approach if REPL.options is implemented.

rfourquet · 2019-10-30T13:35:49Z

I've updated to put the iocontext field with the options one, which contains all user-settable configuration (as opposed to internal fields).

I will merge next time CI is green enough :)

This adds an `iocontext::Dict{Symbol,Any}` field to `LineEditREPL.options`, which can be initialized with `atreplinit`, and updated interactively, e.g. `Base.active_repl.options.iocontext[:compact] = true`. Fixes #20509.

clarkevans · 2020-07-07T18:18:41Z

I would note that this fantastic feature isn't usable in a startup file; probably because the REPL is not initiated yet? What would be a good solution to this is having a second startup file... for REPL only. ERROR: LoadError: UndefVarError: active_repl not defined

KristofferC · 2020-07-07T18:35:39Z

Doesn't this work?

https://github.com/JuliaLang/julia/pull/29249/files#diff-b82b6651ae593ab1e85f79f4f1280493R79-R84

clarkevans · 2020-07-07T19:09:57Z

Right... it's even in the documentation. That said, this probably deserves its own section heading, on par with to setting color and key bindings. Thanks again.

rfourquet added the domain:display and printing Aesthetics and correctness of printed representations of objects. label Sep 18, 2018

rfourquet mentioned this pull request Aug 8, 2019

function to change numeric output format #6493

Open

rfourquet force-pushed the rf/repl/iocontext branch from 6613f8c to 5aa66f5 Compare September 17, 2019 06:38

rfourquet added the stdlib:REPL Julia's REPL (Read Eval Print Loop) label Sep 17, 2019

rfourquet force-pushed the rf/repl/iocontext branch from 5aa66f5 to b386f5f Compare September 25, 2019 15:13

KristofferC reviewed Oct 1, 2019

View reviewed changes

rfourquet force-pushed the rf/repl/iocontext branch 2 times, most recently from 9a720f2 to caf2950 Compare October 2, 2019 11:11

KristofferC mentioned this pull request Oct 7, 2019

Floating-point error converting floats to strings #33485

Closed

rfourquet force-pushed the rf/repl/iocontext branch from caf2950 to f54477e Compare October 30, 2019 13:33

REPL: allow tweaking the implicit IOContext of the REPL

4212cc9

This adds an `iocontext::Dict{Symbol,Any}` field to `LineEditREPL.options`, which can be initialized with `atreplinit`, and updated interactively, e.g. `Base.active_repl.options.iocontext[:compact] = true`. Fixes #20509.

rfourquet force-pushed the rf/repl/iocontext branch from f54477e to 4212cc9 Compare November 1, 2019 17:45

rfourquet mentioned this pull request Nov 1, 2019

[WIP] allow sorting Dict/Set values in show #33744

Open

rfourquet merged commit 9e20e03 into master Nov 16, 2019

rfourquet deleted the rf/repl/iocontext branch November 16, 2019 09:20

mbauman mentioned this pull request Jul 10, 2020

fix #36492, ensure output is limited even with basic REPLs #36610

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

REPL: allow tweaking the implicit IOContext of the REPL #29249

REPL: allow tweaking the implicit IOContext of the REPL #29249

rfourquet commented Sep 18, 2018 •

edited

Loading

rfourquet commented Sep 17, 2019

rfourquet commented Sep 25, 2019

KristofferC commented Sep 25, 2019 •

edited

Loading

rfourquet commented Sep 25, 2019

rfourquet commented Sep 27, 2019

StefanKarpinski commented Oct 1, 2019

KristofferC Oct 1, 2019

rfourquet Oct 2, 2019

KristofferC Oct 2, 2019

rfourquet Oct 2, 2019

KristofferC Oct 2, 2019

KristofferC Oct 2, 2019

rfourquet Oct 2, 2019

KristofferC Oct 2, 2019

rfourquet Oct 2, 2019

tkf Oct 2, 2019

rfourquet commented Oct 30, 2019

clarkevans commented Jul 7, 2020 •

edited

Loading

KristofferC commented Jul 7, 2020

clarkevans commented Jul 7, 2020

REPL: allow tweaking the implicit IOContext of the REPL #29249

REPL: allow tweaking the implicit IOContext of the REPL #29249

Conversation

rfourquet commented Sep 18, 2018 • edited Loading

rfourquet commented Sep 17, 2019

rfourquet commented Sep 25, 2019

KristofferC commented Sep 25, 2019 • edited Loading

rfourquet commented Sep 25, 2019

rfourquet commented Sep 27, 2019

StefanKarpinski commented Oct 1, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rfourquet commented Oct 30, 2019

clarkevans commented Jul 7, 2020 • edited Loading

KristofferC commented Jul 7, 2020

clarkevans commented Jul 7, 2020

rfourquet commented Sep 18, 2018 •

edited

Loading

KristofferC commented Sep 25, 2019 •

edited

Loading

clarkevans commented Jul 7, 2020 •

edited

Loading