Fix duckdb leading comments

[marshallwhiteorg]

Describe your changes

Fix empty result in certain duckdb SELECT and SUMMARIZE queries with leading comments

Issue number

Closes #892

Checklist before requesting a review

• Performed a self-review of my code
• Formatted my code with pkgmt format
• Added tests (when necessary).
• Added docstring documentation and update the changelog (when needed)

---

📚 Documentation preview 📚: https://jupysql--895.org.readthedocs.build/en/895/

[edublancas]

@marshallwhiteorg any updates here?

[marshallwhiteorg]

@edublancas still some issues, will continue in the AM https://github.com/ploomber/jupysql/actions/runs/6320834470/job/17163937282?pr=895

[edublancas]

some of these look like matplotlib issues. I think it got upgraded recently and we had to pin the version:

jupysql/setup.py

Line 47 in cf824ca

"matplotlib==3.7.2",

merging from master might help

[marshallwhiteorg]

@tonykploomber @neelasha23

I have all the tests passing except some of the docs are failing to build. My code is using sqlglot.parse_one to parse the duckdb queries and that function doesn't accept this type of query:

%%sql duckdb:///
INSTALL 'sqlite_scanner';
LOAD 'sqlite_scanner';

The error is in here: https://readthedocs.org/projects/jupysql/builds/22054160/

Here's an example.

In [44]: from sqlglot import parse_one
In [45]: parse_one("INSTALL 'sqlite_scanner'", dialect="duckdb")

...
ParseError: Invalid expression / Unexpected token. Line 1, Col: 24.
  INSTALL 'sqlite_scanner'

Can you see what I am doing wrong, or do you know why this used to work? Is there some preprocessing step that sqlglot requires for this query? It looks like valid duckdb syntax based on their docs. Also, it works with double quotes but not single quotes. Thanks!

[neelasha23]

I can see this query is failing:

%%sql
SELECT 
    name, 
    friends[1] AS first_friend, 
    likes.pizza AS likes_pizza, 
    likes.tacos AS likes_tacos
FROM read_json_auto('people.json')

The issue might be in this line : friends[1] AS first_friend,

It looks like sqlglot had a release yesterday : https://pypi.org/project/sqlglot/#history
You can check the release notes to see if there's any breaking changes. Or you can pin sqlglot version < 18.8.0 and rerun the CI. If it fixes the problem open an issue that the latest sqlglot is causing queries to break. @marshallwhiteorg

[marshallwhiteorg]

@edublancas @idomic @neelasha23 Okay I've made the requested changes, please let me know if you see issues with the new version.

I was skeptical that sqlparse would be up to the task but I was only able to make it fail once where sqlglot succeeded.
For the query SELECT FROM table WHERE (column = 'value' sqlglot correctly marks it as invalid due to the missing close paren but sqlparse does not. I marked the test as xfail. Let me know if you have any more test cases you'd like me to add.

[edublancas]

minor stuff. remember to respond to PR comments with a link to your changes: https://ploomber-contributing.readthedocs.io/en/latest/contributing/responding-pr-review.html

[edublancas]

@marshallwhiteorg addressed the comments. please check the open conversations and request my review when done

[marshallwhiteorg]

@edublancas Thanks for the comments, I just responded to both and made the assert -> NotImplementedError change, see my comment. Ready for review again

[edublancas]

lgtm. @neelasha23 please run some manual tests and let us know if everything looks good

[neelasha23]

lgtm!

[neelasha23]

please resolve merge conflicts @marshallwhiteorg

[edublancas]

@marshallwhiteorg please resolve merge conflicts and let me know so we can merge this

[marshallwhiteorg]

@edublancas @neelasha23 Fixed merge conflicts

[edublancas]

@marshallwhiteorg tests are failing

[marshallwhiteorg]

@edublancas fixed