more complex visit_Call to parse chained command #5677
Conversation
|
Thank you for your pull request! We could not find a changelog entry for this change. For details on how to document a change, see the contributing guide. |
1 similar comment
|
Thank you for your pull request! We could not find a changelog entry for this change. For details on how to document a change, see the contributing guide. |
|
Did some rough benchmark on performance implication. For the the previous case in unittest, visiting the tree 10000 times, time with previous method takes 0.34s, new method takes 0.34s, as there's no additional visit needed. For the test case in unittest, visiting the tree 10000 times, time with previous method takes 0.34s, new method takes 0.71s. With the whole content in run.py, visiting 10000 times takes 7.65s for original method, takes 8.23s for new method. ConclusionThe new method will slow things down but don't seems to cause performance concern. |
| func_name = func_name.split(".")[-1] | ||
| args, kwargs = self._get_call_literals(node) | ||
| self.dbt_function_calls.append((func_name, args, kwargs)) | ||
| if isinstance(node.func, ast.Attribute) and node.func.attr in dbt_function_key_words: |
There was a problem hiding this comment.
@gshank Do you have any experience with language parser? feels like we might be re-inventing the wheels now
|
Another thought for this is that there are tools that can convert regex to a language parser I think @jtcohen6 (the book we are reading mentioned it). Maybe at some point we can look into switch if performance/complex rules become a issue |
* more complex visit_Call * add changelog * traversing all of the tree
* more complex visit_Call * add changelog * traversing all of the tree
* more complex visit_Call * add changelog * traversing all of the tree
resolves #5646
Description
Added more node traversing method to handle new test cases added in the unittest
Checklist
changie newto create a changelog entry