Add read_archive and list_archive_entries support #389
Conversation
This file contains hidden or 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
yongtang
changed the title
This PR is part of the discussion in 366 where we want to reduce the C++ implementation by reusing primitive ops for archive support. The idea is to implement C++ kernel that could be used in general, than adding additional C++ for each Dataset. Summary: 1) read_archive_entries could probe for archive format and returns entry names in archive automatically. 2) read_archive can take the output of read_archive_entries, and returns string tensor as content of the entries in archive. 3) both read_archive_entries and read_archive support in graph and eager mode and could be wired to implement an "ArchiveDataset" if needed (see demo in tests/test_archive_eager.py) 4) combined with other PR 384 it is possible to read memory and decode parquet like file in memory, by utilizing `SizedRandomAccessFile` which is a subclass of tensorflow's `RandomAccessFile`. Overall, we expect files compressed in archive not to be too big so it should at least fit into CPU memory. While files such as HDF5 or parquet could be really large, the compression normally happens at the file format level so additional archive/compression outside with huge file (e.g. hundreds of GBs) are rarely heard of. So, the implementation in this PR should be enough for normal need. Signed-off-by: Yong Tang <[email protected]>
Signed-off-by: Yong Tang <[email protected]>
Signed-off-by: Yong Tang <[email protected]>
Signed-off-by: Yong Tang <[email protected]>
|
/cc @terrytangyuan |
terrytangyuan
approved these changes
Aug 6, 2019
i-ony
pushed a commit
to i-ony/io
that referenced
this pull request
Feb 8, 2021
* Add read_archive and read_archive_entries support This PR is part of the discussion in 366 where we want to reduce the C++ implementation by reusing primitive ops for archive support. The idea is to implement C++ kernel that could be used in general, than adding additional C++ for each Dataset. Summary: 1) read_archive_entries could probe for archive format and returns entry names in archive automatically. 2) read_archive can take the output of read_archive_entries, and returns string tensor as content of the entries in archive. 3) both read_archive_entries and read_archive support in graph and eager mode and could be wired to implement an "ArchiveDataset" if needed (see demo in tests/test_archive_eager.py) 4) combined with other PR 384 it is possible to read memory and decode parquet like file in memory, by utilizing `SizedRandomAccessFile` which is a subclass of tensorflow's `RandomAccessFile`. Overall, we expect files compressed in archive not to be too big so it should at least fit into CPU memory. While files such as HDF5 or parquet could be really large, the compression normally happens at the file format level so additional archive/compression outside with huge file (e.g. hundreds of GBs) are rarely heard of. So, the implementation in this PR should be enough for normal need. Signed-off-by: Yong Tang <[email protected]> * Fix macOS build failure Signed-off-by: Yong Tang <[email protected]> * Fix 1.14 build failure Signed-off-by: Yong Tang <[email protected]> * Rename to list_archive_entries and read_archive Signed-off-by: Yong Tang <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
This PR is part of the discussion in #366 where we want to reduce the C++ implementation by reusing primitive ops for archive support.
The idea is to implement C++ kernel that could be used in general, than adding additional C++ for each Dataset.
Summary:
SizedRandomAccessFilewhich is a subclass of tensorflow'sRandomAccessFile.Overall, we expect files compressed in archive not to be too big so it should at least fit into CPU memory. While files such as HDF5 or parquet could be really large, the compression normally happens at the file format level so additional archive/compression outside with huge file (e.g. hundreds of GBs) are rarely heard of. So, the implementation in this PR should be enough for normal need.
Signed-off-by: Yong Tang [email protected]