Fix syntax highlight for pgsql mode for backslash and escape constant

[adityatoshniwal]

Fix syntax highlight for pgsql mode as PostgreSQL does not require backslash to be escaped.
Add support for escape constant as mentioned with ref to:
https://www.postgresql.org/docs/current/sql-syntax-lexical.html#SQL-SYNTAX-STRINGS-ESCAPE

Below are the sample queries to test:
SELECT 'select c:\data\out'||oid||'.csv' from pg_class;
select E''', '123';

Attached are the screenshots of working example where the above queries works fine - pgAdmin4.

[marijnh]

It looks like you are keeping mode state in a variable, if I understand the role of escapeConstSet correctly. That doesn't work—when the mode is resumed from some random point in the code, that state will still be leftover from the last time it ran, and things will break.

[adityatoshniwal]

Urgh.. Right. Do you have any suggestions on how I can improve it ?

[marijnh]

You could put it in the state object instead.

[adityatoshniwal]

OK, thanks. I'll go through the code once again.

[adityatoshniwal]

@marijnh I've updated the code. Kindly review.

[marijnh]

Two more questions:

• Is it intentional that this is also being set for other types of string prefixes like n and _? It looks like it should only be relevant for e.

• You don't actually need a new state field, since the state already has a tokenize callback that you could replace with a tokenizer function that can parse escaped strings. You could add another argument to tokenLiteral that determines backslash behavior, and set tokenize to a function that consumes the quote and then replaces itself with tokenLiteral(quote, ...) here.

[adityatoshniwal]

Two more questions:

• Is it intentional that this is also being set for other types of string prefixes like n and _? It looks like it should only be relevant for e.
  Oh yes, you're right. Will fix this. :/
• You don't actually need a new state field, since the state already has a tokenize callback that you could replace with a tokenizer function that can parse escaped strings. You could add another argument to tokenLiteral that determines backslash behavior, and set tokenize to a function that consumes the quote and then replaces itself with tokenLiteral(quote, ...) here.
  I didn't want to repeat the logic of tokenLiteral and wanted to reuse it. I'll try if can avoid the new state field.

Thanks for your time.

[marijnh]

I didn't want to repeat the logic of tokenLiteral and wanted to reuse it.

That's what I mean—if you give it an extra argument, you can reuse it.

[adityatoshniwal]

Cool. Let me try it !! Thanks.

[adityatoshniwal]

Hi,
I tried to do it without escapeConstSet var, but the problem is, it is before the string stream starts. So I have to tell the tokenLiteral function somehow that the previous char was a escape constant. I can do it in escape constant check, but then I lose the formatting for escape constatnt.

[marijnh]

After seeing the e prefix, you can set state.tokenize to a tokenizer function that's specific to that situation, which starts tokenizing an escape-supporting string when it is called.

[adityatoshniwal]

Hi @marijnh,
This looks a lot better now. :)
Sorry for trouble.

[marijnh]

Thanks for implementing this. I've slightly adjusted it further in attached patch, so that the code won't try to skip quotes at the start of continued string lines.

[adityatoshniwal]

Thanks a lot :)