Proposal: DrRacket Surrogates for meta-languages.

[Leif Andersen]

BACKGROUND:

DrRacket uses a language's get-info function to set parameters such as
syntax coloring, indentation, and toolbar buttons. One parameter is
`definitions-text-surrogate`[1], which lets `#lang`s set a surrogate
for DrRacket. The docs mention that this is a powerful, yet easy to do
wrong, IDE extension mechanism.

Unfortunately, as is, the `definitions-text-surrogate` form does not
mix well with meta-languages e.g. s-exp, reader, at-exp. When given
the `definitions-text-surrogate` key, get-info returns a single module
path for DrRacket to `dynamic-require`. There is no way for a
meta-language to mix in its surrogate with the surrogate of the
language its extending. (At least not without either creating a
temporary file on the fly, or trying to scrape the current state of
DrRacket. Both of which are terrible ideas.)




PROPOSAL:

We extend the valid values for the `definitions-text-surrogate` key to
be a list of symbols as well as a single symbol. In the case of a
single symbol, nothing changes.

In the case of a list, each element must also be a module path. The
last element must refer to a module that provides a `surrogate%` class
as before. Every other element, however, refers to a module that
provides a `surrogate%` _mixin_. The IDE constructs the expected
surrogate by traversing the list, each time `dynamic-require`ing the
module, and mixing in the result. Meta-languages can augment existing
surrogates, rather than having to throw them away entirely. The
`get-info` code for a meta-language might look something like this:

```
(make-meta-reader
...
(lambda (base-get-info)
(lambda (key default)
(case key
[(definitions-text-surrogate) (flatten (list
"meta-surrogate.rkt" (base-get-info key default)))])))
```



ENGINEERING EFFORT:
It looks like all of the changes would be confined to the following files:
drracket/drracket/drracket/private/in-irl-namespace.rkt
drracket/drracket/drracket/private/insulated-read-language.rkt
drracket/drracket/drracket/private/module-language-tools.rkt

(Obviously also documentation+tests)

[1]: http://docs.racket-lang.org/tools/lang-languages-customization.html?q=Definitions%20Text%20Surrogate#%28part._sec~3adefinitions-text-surrogate%29

--
~Leif Andersen

[Robby Findler]

Except for the fact that the definitions-text-surrogate was kind of dubious to begin with, this sounds like a good idea to me!

Robby

--
You received this message because you are subscribed to the Google Groups "Racket Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to racket-dev+...@googlegroups.com.
To post to this group, send email to racke...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/racket-dev/CAAVaeECiQhKiziKVp6wOOWm6vQM1GPr90yPAOHoyOHDBJ3c62g%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

[Leif Andersen]

> Can we put this into the relevant repo (drracket) please?

Yup. I've opened a issue in the DrRacket repo, and am working on a PR now.

> I think Robby gave you the go-ahead.

That's the impression I got too.

> Except for the fact that the definitions-text-surrogate was kind of dubious to begin with, this sounds like a good idea to me!

Ha...ya...that it is. Hopefully we can eventually make something
similar in power, but safer soon. ;)

On 12/2/18, Matthias Felleisen <matt...@felleisen.org> wrote:
>
> Can we put this into the relevant repo (drracket) please?
>
> I think Robby gave you the go-ahead. — Matthias

>> --
>> You received this message because you are subscribed to the Google Groups
>> "Racket Developers" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to racket-dev+...@googlegroups.com.
>> To post to this group, send email to racke...@googlegroups.com.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/racket-dev/CAAVaeECiQhKiziKVp6wOOWm6vQM1GPr90yPAOHoyOHDBJ3c62g%40mail.gmail.com.
>> For more options, visit https://groups.google.com/d/optout.
>
>

--
~Leif Andersen