Extended support for function shorthand notation

[maxecharel]

Adds font-locking support for R function shorthand notation \() and for names of functions defined using this shorthand notation. Also makes the shorthand notation a recognized function pattern.

[lionel-]

Would you mind taking a look and add to the existing unit tests for font-locking in https://github.com/emacs-ess/ESS/blob/master/test/ess-test-r-fontification.el? If you run into trouble please let me know.

[lionel-]

I'm not sure we want to encourage such usage but I guess it makes sense to have a baseline support and let linters/reformatters correct this usage.

[maxecharel]

Personally I only use the shorthand notation for lambdas, and I agree with you but since the options has been made available by R Core, I think it is a good thing to have it supported by ESS. I see your point concerning the 'encouragement', but I do not consider this as a real one (including snippet templates would have been a different story^^)

[maxecharel]

Hi @lionel- , sorry for the delay before answer. I will add to the existing unit tests for font-locking end of this week. I will get back to you. Thanks for the great work with ESS !

[maxecharel]

Hi @lionel- , your feedback would be welcome concerning the two following issues I encountered when adding to the unit tests:

1. while \() passes the ess-test-r-fontification-keywords-fn-test test, I had to go for a 'trial and error' approach that I describe below. To investigate the reasons why it failed in the first place, I would like to get access to the ERT test buffer (not just the results) to see how the string is printed. Any idea how to do so?
2. when function name is backquoted, it is not highlighted (I realized that when trying to add to ess-test-r-fontification-keywords-disabled-backquoted-defun), i.e.:
   
   I failed to identify the problem. I will carry on investigating in the following days, but maybe you have some hints on the nature of the issue?

Concerning (1), at first ess-test-r-fontification-keywords-fn-test failed. It was rather surprising given that manually executing (face-at-point) in my minimal Emacs testing config worked as expected:

Anyway, ERT indicated a failure whenever the relevant string was set (in the :case string) to \\\\(), \\() or even \()
(with the pilcrow character located at the relevant place depending on :case or :result).

My first guess was that the problem comes from the way ERT, based on the :case of ess-test-r-fontification-keywords-fn-test, prints the string representing \() in a test buffer. After some trials and errors, I ended up using group delimiters --- i.e. I went for \\(\\\\\\)() --- and it passed the test. Yet I am not satisfied with this, and I would like to get a deeper understanding of what happened exactly. Is there any way I can get access to the actual test buffer in which the expressions contained in :case are printed and actually tested? Or am I missing the point and there is no creation of such buffer?

Thanks.
Best,
Max

[lionel-]

@maxecharel I will try to find time to take a look this week.

[maxecharel]

@lionel- some updates (to keep you posted and as notes for myself):

1. I understood why backquoted 'names' of functions defined with \() were not fontified; solved with commit 5c24a72 by modifying ess-r-font-lock-syntactic-face-function
   
2. I discovered an issue with my initial modification of ess-R-function-name-regexp while I was investigating the previous issue and fixed it with commit 49bcd58:
   
   

One question remains, and one has been added (this one will certainly require your feedback at some point, sorry for that)

1. how to get access to the ERT test buffer (nothing new here; I'll try to answer that later this week while reruning the tests, but if and only if you find time, your feedback is more than welcome)
2. definitions of ess-r-mode and ess-r-transcript-mode (as well as inferior-ess-r-mode) all reference ess-r-font-lock-syntactic-face-function, but they (except the latter mode) also set a local add-log-current-defun-header-regexp including a string in the spirit of "<- function". However, here it looks like we enter a world that goes beyond font-locking (and therefore the scope of this PR). When you have the opportunity, could you just confirm this? Anyway I will set that aside for now, and will try to find the time to dig into it later.
   Cheers!

[lionel-]

Sorry I don't have any bandwidth for this :/

Let's just merge your work.