| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
it's a bit of a mess and should be using the syntax reader instead, but not
sure if it's worth it.
|
|
same code that finds putative definitions, with all its caveats
|
|
i think we've been sending twice the symbol at point in ge:autodoc's list of
paths since the beginning of time; not that it'll make a noticeable
difference, but let's not.
|
|
we already had our own lighter mechanism, just needed to use it better. it
will also allow guessing local signatures, quite useful in chezzy (or more
generally r6rs-librarish) schemes.
|
|
|
|
|
|
see discussion in issue #56
|
|
|
|
|
|
we were doing it really wrong: first, by not setting the evaluation
environment when things are computed outside the original buffer (the
completion in region case for consult and the likes, where the current buffer
is the minibuffer) and, to add insult to injury, by not waiting for a response
from the scheme side!
actually waiting can make things a tad laggy when the scheme is not as fast
as, say, chez and there are lots of completions (if one uses
completion-in-region; for pop ups like company it's fine), so i've also added
a flag, geiser-autodoc-use-docsig, to turn the functionality off.
|
|
... as well as a way of telling imenu to look for nested define forms, as the
ones one finds for instance inside (library ...) or (module ...) sexps, or
simply nested defines in function bodies. it's a crappy way of finding
definitions, but it's better than nothing when it's all we have (e.g., R6RS
libraries don't seem to provide an environment/namespace including their
privates, which is a killjoy).
|
|
Chez likes to do that, which is fine now.
|
|
We were calling an undefined function... not too many people using the
functionality (or geiser at all :)).
|
|
... even if that protocol is unwritten (note to self: write it already) and
nobody till now was being a good citizen (chez is becoming one).
|
|
|
|
- geiser-log: Obsoleting some -p flags and lexical binding
- geiser-eval: Ensuring logs for asynchronous retorts
- geiser-debug: don't bail on no output with errors
|
|
Should address issue #55 for all implementations.
|
|
Thanks to Panagiotis Vlantis, see issue #54
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Some authors of downstream packages are reluctant to do stuff with
upstream packages' private functions.
(for example, https://github.com/abo-abo/swiper/pull/2235#issuecomment-537399457)
|
|
Should address, sort of, issue #46
|
|
|
|
Thanks to Brian Leung
|
|
|
|
|
|
|
|
Should fix issue #49.
|
|
See discussion in issue #48.
|
|
Should fix issue #48
|
|
|
|
|
|
|
|
NOTE: The patch is largely untested.
Modifications:
- Update readme.org
- Remove geiser-company
- Move Company extensions to geiser-completion
Omissions:
- geiser-company--inhibit-autodoc has been removed. Eldoc handling
should be implemented in the frontend, not in the backend.
See for example:
https://github.com/minad/corfu/blob/04fbfce3d7e9c125a7fd22a34455a508247a522b/corfu.el#L1212
- The quickhelp-string action and geiser-company--docstring have been
removed. company-quickhelp can use `:company-doc-buffer` instead with
minimal overhead.
See:
https://github.com/company-mode/company-quickhelp/blob/3ca2708b4e5190205aca01d65fe1b391963a53f9/company-quickhelp.el#L138
- The automatic Company setup has been removed. Personally I am not a
fan of such auto configuration. It is better if completion is
configured consistently in the user configuration. You may want to
restore the auto configuration for backward compatibility. It depends
on your backward compatibility story. I am fine with rare breaking
changes from time to time.
- There is a cyclic dependency between geiser-edit/geiser-doc and
geiser-completion, which should be untangled.
|
|
And, actually, fixing bugs: the warnings weren't spurious.
|
|
Playing the trick of checking for eldoc-documentation-functions to
know wheter we're at a new enough version, and (hopefully) falling
back to the old implementation otherwise.
|
|
We remove by-now obsolete usage of eldoc, and depend on the elpa
package to ensure backwards compatibility.
|
|
Fixes #45. Thanks to Brian Cully.
|
|
The byte-compiler complained about it.
|
