Add Locking to TableInfo to Support Mutating a DataFrame from Multiple Threads
#427
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ainst the classification labels
deserialize stage to slice up the input and trigger the bug
…us into david-add-scores-bug
…etection stage tests, also fix multi-segment tests
mdemoret-nv
added
breaking
Closed
|
Moving to 23.01 since the majority of the multi-threaded issues have been addressed. |
|
|
||
| // return py_table; | ||
| return self.get_py_table(); | ||
| return self.get_info().copy_to_py_object(); |
There was a problem hiding this comment.
If we make this change, we should also make a copy in the Python impl of this class.
Otherwise we will get different behaviors in Python stages when C++ execution is enabled/disabled.
|
Closing in favor of #586 |
ghost
pushed a commit
that referenced
this pull request
Feb 6, 2023
* Builds on changes in #427 * Adds a `PreallocatorMixin` which when added to a stage performs pre-allocation. This should be added to the first stage in a pipeline which emits a DataFrame or MessageMeta in a pipeline. * Morpheus' TypeId enum exposed to the Python API, allowing stages to define types for columns needing pre-allocation * `MutableTableInfo` exposed to Python via a context manager to be used in `with` blocks * `type_util` (`Dtype`) and `type_util_detail` (`DataType`) merged into a new compilation unit `dtype` fixes #490 fixes #456 Authors: - David Gardner (https://github.com/dagardner-nv) - Michael Demoret (https://github.com/mdemoret-nv) Approvers: - Michael Demoret (https://github.com/mdemoret-nv) URL: #586
jjacobelli
pushed a commit
to jjacobelli/Morpheus
that referenced
this pull request
Mar 7, 2023
* Builds on changes in nv-morpheus#427 * Adds a `PreallocatorMixin` which when added to a stage performs pre-allocation. This should be added to the first stage in a pipeline which emits a DataFrame or MessageMeta in a pipeline. * Morpheus' TypeId enum exposed to the Python API, allowing stages to define types for columns needing pre-allocation * `MutableTableInfo` exposed to Python via a context manager to be used in `with` blocks * `type_util` (`Dtype`) and `type_util_detail` (`DataType`) merged into a new compilation unit `dtype` fixes nv-morpheus#490 fixes nv-morpheus#456 Authors: - David Gardner (https://github.com/dagardner-nv) - Michael Demoret (https://github.com/mdemoret-nv) Approvers: - Michael Demoret (https://github.com/mdemoret-nv) URL: nv-morpheus#586
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
If you run multiple threads with the
AddScoresorAddClassificationstages, they may try to append a new column to the dataframe. In this situation, you can have race conditions where some threads are using the dataframe object at the same time you are adding/removing columns.To fix this, this PR separates the
TableInfoclass into two classes:TableInfo2. Allow for read operations to happen in parallel
MutableTableInfo4. Requires exclusive access to the underlying dataframe before it can be constructed. Allows for add/remove operations on columns
The effect is that holding an instance of
TableInfowill ensure that the column types and count do not change for the lifetime of that object. Holding an instance ofMutableTableInfoensures that you have exclusive access to the dataframe and can mutate it.Outstanding questions:
as_py_object()method onTableInfo? This should probably make a copy.