Draft flow PR #127
Draft flow PR #127
Conversation
| if not files_without_tests: | ||
| raise Exception("No java files were found") | ||
|
|
||
| full_dataset_folder = Path(self.dir_to_save) / 'full_dataset' |
There was a problem hiding this comment.
long method. the part with directory handling could be extracted, for example
| for filename in tqdm(files_without_tests): | ||
| try: | ||
| text = next(result) | ||
| if text: |
There was a problem hiding this comment.
under what condition can a text be None?
There was a problem hiding this comment.
When result is empty, empty string, when the file consists only of comments
There was a problem hiding this comment.
when can a result be empty, beside the case when soure code is empty ?
There was a problem hiding this comment.
when we pass empty file, i believe.
There was a problem hiding this comment.
can u leave a comment about it, maybe?
| csv = self.inputLoad()['csv'] | ||
| rows = [x for _, x in csv.iterrows()] | ||
|
|
||
| with Parallel(n_jobs=2, require='sharedmem') as parallel: |
There was a problem hiding this comment.
it's just a template, we can fix in future
There was a problem hiding this comment.
can u use self.system_cores_qty, like u did in the preprocess task?
|
|
||
| with Parallel(n_jobs=2, require='sharedmem') as parallel: | ||
| results = parallel((delayed(self._find_EMs)(a) for a in rows)) | ||
| self.save({"data": [x for x in results if x]}) |
There was a problem hiding this comment.
the result of _find_EMs function
There was a problem hiding this comment.
can u type-annotate the _find_EMs function then?
| def _find_EMs(self, row): | ||
| result_dict = {} | ||
| try: | ||
| ast = AST.build_from_javalang(build_ast(row['original_filename'])) |
There was a problem hiding this comment.
long method. this whole routine about extracting method declrations could be put into a separate method
| target_node = ast.get_subtree(method_node) | ||
| for method_invoked in target_node.get_proxy_nodes( | ||
| ASTNodeType.METHOD_INVOCATION): | ||
| extracted_m_decl = method_declarations.get(method_invoked.member, []) |
There was a problem hiding this comment.
what is extracted_m_decl ? what is its type?
| dir_to_save = d6tflow.Parameter() | ||
| system_cores_qty = d6tflow.IntParameter() | ||
|
|
||
| def _find_EMs(self, row): |
There was a problem hiding this comment.
can u rename row to a bit more informative? what's in the row?
| dir_to_save=args.output, | ||
| system_cores_qty=args.jobs | ||
| )) | ||
| data = TaskFindEM( |
There was a problem hiding this comment.
task TaskFindEM doesn't look like the final step. or is it?
where do u write the final dataset to disk?
d6tflow can't run in parallel child tasks
it can save results in a task only 1 time (child task can be only run synchronously). So if the task saves smth, it is completed. We have to collect all results from child tasks and then pass large collection to the next task. And do it every time from task to task
Since we have to pass large collection we will keep in memory files and it's metadata.
We can keep only metadata of all files, and then open each files in a separate thread and then find EM, Target method and it's invocation. But it is very ugly and take too much time
I have to use
joblibsinceProcessPoolcannot serialize some AST objectsYou can run program as
python3 main.py --dir /mnt/d/temp/dataset/01/ --output output