Ensure we output transpiled code

[gueniai]

What does this PR do?

Currently, if there's a parsing error (usually, these mean the output is unusable) we copy the input code into the output file. This leads to a confusing interaction, because there is no indication that the file is the same as the input, and the user is left wondering what happened.

This PR changes the output, so that the transpiled code, including all issues we found, are available in the "transpiled" file.

Tests

• manually tested
• added unit tests
• added integration tests

[github-actions]

✅ 27/27 passed, 1 flaky, 1m20s total

Flaky tests:

• 🤪 test_transpile_sql_file (14.47s)

_{Running from acceptance #2340}

[vil1]

I don't think this is the proper way to go here.

When there is some parsing error, morpheus doesn't currently guarantee it won't "cut" some of its input.

For instance, when attempting to transpile

ADD SIGNATURE TO ProcForAlice BY CERTIFICATE csSelectT

which uses syntax that isn't currently covered by the grammar, morpheus will output the following

/* Parse error: unexpected extra input 'ADD' while parsing a sqlFile
expecting one of: @Local, End of batch, Identifier, Node ID, Select Statement, Statement, '(', ';', 'ARRAY', 'BULK', 'CALL', 'DBCC'...
'TO' was unexpected while parsing a sqlFile
expecting one of: End of batch, Select Statement, Statement, '(', 'BULK', 'CALL', 'COMMENT', 'CONTRACT', 'DBCC', 'DISABLE', 'ENABLE', 'GET'... */
-- FIXME: Unparsed input - ErrorNode encountered
;
/* SIGNATURE */
-- FIXME: SNOWFLAKE: The transpiler cannnot currently convert Execute body batch, but may be able to do so in the future
;

where most of the input query has disappeared (in this instance, we've lost everything but the SIGNATURE token).

What I suggest we do instead is to keep outputting context.source_code but we decorate it with comments listing the parsing errors encountered (the code for doing that should still be present somewhere).

[gueniai]

@vil1 I do like your idea more as the long-term solution, although we need to think about how we would handle the output from the BB converter. I still think this initial change is a step in the right direction. Outputting the input source with 0 context is very disorienting.

[sundarshankar89]

LGTM, may be an additional tests for this behavoiur will show an example of how morpheus behaves for failures.

[vil1]

Ready to approve once my little comment is addressed and https://github.com/databrickslabs/morpheus/pull/509 is merged

[vil1]

we should remove the whole if