[SQL] Many fixes and improvements overall

[azizk]

Hey!

My main goal was to fix the way the word-boundary escape \b is used throughout the syntax file. I did much more besides that.
So I wanted, for example, 1IN to be two tokens, and I assumed that every engine recognizes them as such, but I just tested with PSQL and MySQL and it looks like it's 2 tokens for PSQL and 1 for MySQL.

-- This is valid in PSQL and returns `true`. MySQL thinks it's an undefined function called "1IN".
SELECT 1IN(1,2,3);

I haven't looked at what the standard says yet. Regardless, I think my changes are pretty useful as they are and the thing above can be reverted by adding a \b back to the match: \d+ context. My last commit is solely for formatting. Hope you like it! 👍

[deathaxe]

Many thanks to your contribution. I wonder whether we can incooperate any valuable changes into #3046 as it already comes a long with a lot of improvements especially with regards to syntaxes for different SQL dialects based on one "root-syntax".

[azizk]

Oh dear, I should have done a search before doing this. 😄
I'll check it out later...

[jrappen]

Yes, as @deathaxe said, please take a look at @keith-hall 's #3046. As you can see, his branch for that PR is https://github.com/forkeith/Packages/tree/sql which you can use for a diff and/or possible PRs to target and be included in #3046.

[azizk]

Thanks @jrappen! Yeah I checked it out and it's looking great. There are some things I'd do differently (the leading \b) and I could also contribute some more stuff, which isn't included in this PR (mostly PostgreSQL). Hey @keith-hall, if you're interested I could contribute and send a PR to your branch. 👍

Nevertheless, do you all think it's still worthwhile to merge this in? I don't mind if it's overwritten later by Keith's work, but until then these changes may be useful and deployed to users earlier? 😃

[keith-hall]

a PR to my branch improving PostgreSQL would be very welcome :)
I'm not sure I agree with the blanket removal of leading \b though. It would make sense to do this for the PSQL variant, sure, but it'd be nice to keep it for the others, so even though merging this PR early could be beneficial for PSQL users, it may be a regression for all others.

[azizk]

@keith-hall Alright, gonna send some commits then. 😃
So regarding the leading word anchor \b, afaics it's not necessary. It's only needed if you don't consume all the word characters (the way it is now), but if you do then using a leading \b everywhere is pretty redundant. Or maybe I'm overlooking something? And in case you want to recognise something like 1IN as a single token, like I described in my PR comment, then adding a trailing \b to the number regexp is all there is to make it work.

[jrappen]

What's the current status with this PR and the request to add your changes to Keith's PR?

So far you have only renamed some stuff on Keith's branch and not added any meaningful changes from this PR to his.

[azizk]

@jrappen From my end the PR here can go in. Keith is concerned about the leading \b however. I asked about it in my previous comment...
I know I only contributed minor things on his branch. I haven't committed my other progress yet. Just had other priorities in the meantime.

[azizk]

I just rebased the branch. If the PR is still useful please merge, otherwise let's close it. I'm lacking the time and resources to continue with this...

[jrappen]

⚠️ Not sure why changing the pops but not updating to sublime syntax v2 didn't break the tests.