VHDL: Add private, view keywords; Distinguish attribute from keyword

[hoonweiting]

I'm not familiar with the language, but based on the file history, I presume that any changes to the language after 2015 are not reflected. There's been at least one new version since then (VHDL-2019, some highlights summarised on another site), and brings at least two new keywords: private and view. I'd confirm with the language reference manual but it costs like $500 and it's not quite worth the cost for what I'll use it for :P

I've also separated the attributes from the keywords, and used a generic pattern instead of a name list for the attributes (similar to Ada).

[github-actions]

JS File Size Changes (gzipped)

A total of 1 files have changed, with a combined diff of -122 B (-15.4%).

file	master	pull	size diff	% diff
components/prism-vhdl.min.js	794 B	672 B	-122 B	-15.4%

Generated by 🚫 dangerJS against 6eab90a

[RunDevelopment]

The coverage test failing isn't your fault.

That test uses a few different strategies to determine whether a pattern has been covered. The most basic strategy is to simply check that a regex produced some token. However, it uses a different strategy for keyword lists. It requires all keyword in the list to procure a token.

The test now fails because you simplified the keyword enough for the coverage test to recognize it as a keyword list, and now it wants you to test all of them.

I'm sorry about that, but could you please add the missing keywords to the test?

but it costs like $500

lol. Yeah, that's not quite worth it.

I've also separated the attributes from the keywords, and used a generic pattern instead of a name list for the attributes

Sounds good.

[hoonweiting]

Ooh that's interesting! I'll update the tests shortly. My apologies for not running all the tests locally!

[hoonweiting]

Ah hmm... I did not realise that library and use are also contained the in keyword list, I thought they were only considered constants.

I'll update the keyword test, but I'm not sure whether to remove constant, or allow library and use to be assigned both constant and keyword, as they are both standard tokens. I'm leaning towards removing constant, but I wonder why those two are constants as well. I might have to look for whatever VHDL docs I can find about it.

[RunDevelopment]

Ooh that's interesting! I'll update the tests shortly. My apologies for not running all the tests locally!

It's not your fault. The coverage check isn't part of the test suite (because it doesn't test the behavior and correctness of anything), so it's hard to find.

library and use

I think they're both keyword. The spec from 2000 describes a library clause (section 11.2, page 153) and a use clause (section 10.4, page 150). The example from page 11 also highlights them as keywords:

[hoonweiting]

@RunDevelopment I see! Thanks for searching that up :)

I ran the test on my laptop, it timed out the first time, but all of it passed after plugging my laptop in.

[RunDevelopment]

Thank you!

I ran the test on my laptop, it timed out the first time

Do you remember which one timed out?

[hoonweiting]

Ah no, not really. I just tried to run it again after unplugging it, but it hasn't failed again yet.

I tried to google some of the words I remember seeing in the error, this question on StackOverflow seems to match it.

As for the file, I believe it was coverage.js, which makes sense because I only ran npm run regex-coverage.

[RunDevelopment]

A timeout on coverage? Huh, that's strange. The test just runs all languages tests, so it doesn't do any heavy computations. Anyway, I'll keep that in mind for the future.

[hoonweiting]

It's probably just me not plugging in my laptop. For example, if I load a game without plugging it in first, it'll be really, really laggy. I also notice this lag when running commands from the terminal. It's probably fine normally!