Support error line/column number reporting during parsing

[lmatz]

Considering it can make SQL debugging quite painful, especially when SQL is long, it is critical to user experience.

Wonder if chatGPT can do a good job of finding the location of this type of error. 🤣

[kwannoel]

+1 for this, I guess we will first persist it in AST and eventually in bound query for error reporting?

Are there any user facing errors that are only reported during optimization? Yes, stream nested loop join etc... but I guess these are rare, don't need to persist line/column number in planner for error reporting.

[BugenZhao]

This requires a lot of work as the sqlparser-rs does not involve any "span"s during parsing. FYI, databend has rebuilt the parser from scratch, which provides enough information for clear error reporting in both parser and planner.

https://databend.rs/blog/databend-release-v0.8#new-parser-the-best-parser
https://databend.rs/blog/new-planner

[fuyufjh]

Refactoring sqlparser-rs may take lots of effort. How about starting by reporting line/pos in the parser now?

[xxchan]

Refactoring sqlparser-rs may take lots of effort. How about starting by reporting line/pos in the parser now?

What do you mean? I think it is reporting line/pos in the parser that requires a large refactoring 🥵

[xxchan]

For reference, some upstream issues:

• Add support of locations in AST sqlparser-rs/sqlparser-rs#524
• Plumb location info through the parser sqlparser-rs/sqlparser-rs#757

[fuyufjh]

@chenzl25 Ping, any updates?

[chenzl25]

@chenzl25 Ping, any updates?

No updates currently.