[GitHub] [arrow] mrkn commented on pull request #7477: ARROW-4221: [C++][Python] Add canonical flag in COO sparse index

2020-07-09 Thread GitBox
mrkn commented on pull request #7477: URL: https://github.com/apache/arrow/pull/7477#issuecomment-656481697 @wesm I've finished rebase and adding tests for zero tensor cases. Please review this and then merge it if possible. ---

[GitHub] [arrow] mrkn commented on pull request #7477: ARROW-4221: [C++][Python] Add canonical flag in COO sparse index

2020-06-30 Thread GitBox
mrkn commented on pull request #7477: URL: https://github.com/apache/arrow/pull/7477#issuecomment-652124757 @pitrou In summary, the constraint of the indices order of SparseCOOIndex was removed, but the new flag field is introduced to state whether or not the indices tensor is ordered in a

[GitHub] [arrow] mrkn commented on pull request #7477: ARROW-4221: [C++][Python] Add canonical flag in COO sparse index

2020-06-26 Thread GitBox
mrkn commented on pull request #7477: URL: https://github.com/apache/arrow/pull/7477#issuecomment-650148002 > > Without the canonical flag, we need to make a copy and sort the data of non-canonical sparse tensor when serializing it because the current SparseCOOIndex has the constraint that

[GitHub] [arrow] mrkn commented on pull request #7477: ARROW-4221: [C++][Python] Add canonical flag in COO sparse index

2020-06-26 Thread GitBox
mrkn commented on pull request #7477: URL: https://github.com/apache/arrow/pull/7477#issuecomment-650089924 @pitrou Without the canonical flag, we need to make a copy and sort the data of non-canonical sparse tensor when serializing it because the current SparseCOOIndex has the constraint

[GitHub] [arrow] mrkn commented on pull request #7477: ARROW-4221: [C++][Python] Add canonical flag in COO sparse index

2020-06-22 Thread GitBox
mrkn commented on pull request #7477: URL: https://github.com/apache/arrow/pull/7477#issuecomment-647860707 > I don't understand the PR's motivation. You say that: > > > To support the integration with scipy.sparse.coo_matrix, it is necessary to add a flag in SparseCOOIndex > > bu