precaptureStackTraces does not enhance stack traces for common errors.

[radicaled]

For common mistakes, like a malformed AQL query via db.query, or an operation like db.save on an edge collection missing _to or _from, the stack traces arangojs generates aren't coherent even with precaptureStackTraces.

I think that's because error responses from the server don't have the enhanced stack traces added to them? EG

arangojs/src/connection.ts

Line 900 in 8fe4682

reject(new ArangoError(res));

These kinds of errors account for almost all the errors I see, so it would be great if precaptureStackTraces enhanced these ones as well.

[radicaled]

Some minor experimentation indicates that modifying

arangojs/src/connection.ts

Line 900 in 8fe4682

reject(new ArangoError(res));

to append task.stack to the ArangoError used for rejecting the promise improves error diagnostics considerably.

Would a PR that appends stack traces generated by precaptureStackTraces to these kinds of errors be accepted?

[pluma]

Looking at this code now, I'm not sure when the stack traces even get added. I'm pretty sure the intention was to also augment the stack traces of HttpErrors and ArangoErrors but I think they're currently bypassing that logic entirely so this probably only affects network errors or something?

I'll look into this.

[radicaled]

... but I think they're currently bypassing that logic entirely so this probably only affects network errors or something?

I believe this is correct: I've only ever seen precaptureStackTrace seem to work when I can't connect to my local ArangoDB instance because I haven't brought it up yet. Currently we're using a framework to capture and enhance ArangoDB errors locally via Error.captureStackTrace + dumping any provided AQL queries into the exception text, but it'd be great to see precatureStackTrace working across more error situations so we can drop that part from our codebase.