Fix high memory usage regression due to namespaces - #132
Open
alx75 wants to merge 1 commit into
Open
Conversation
This fixes the high memory usage regression that was introduced in 2.6.2
|
@pythonicrubyist is there any reason we can't get this one merged? |
michaelbaudino
added a commit
to aktio-cc/creek
that referenced
this pull request
Aug 10, 2026
`Reader#namespaces` and `Reader#attributes` both go through `xmlTextReaderExpand()`, which materialises the subtree under the current node. The prefix detection runs on every node until it latches, so it runs on the worksheet root -- where "the subtree under the current node" is the entire document. One call therefore builds a DOM of the whole worksheet and the streaming parse is lost before the first row is yielded. The cost is ~25x the size of the sheet XML, in native memory: 1.9 MB -> 79 MB, 7.4 MB -> 220 MB, 22.7 MB -> 592 MB, 44.8 MB -> 1120 MB. Being native, it appears in neither `GC.stat` nor allocation profiles, which is why it survived several rounds of allocation-focused tuning: on a 44.8 MB sheet, pythonicrubyist#138 alone cuts wall clock in half but leaves RSS at 1149 MB, and pythonicrubyist#132 brings RSS down to 113 MB but doubles wall clock by calling `namespace_uri` on every node and keeping `attributes` on every cell. `Reader#prefix` and `Reader#namespace_uri` map to `xmlTextReaderConstPrefix` and `xmlTextReaderConstNamespaceUri`, which read the current node only. Taking the prefix from the first node that is *in* the spreadsheetml namespace, rather than from the namespaces declared *on* it, also handles a worksheet whose root is in another namespace while its rows are prefixed. Measured on a real 57895-row FEC export (44.8 MB of sheet XML, 5 MB zipped), on top of pythonicrubyist#138: 1149 MB -> 113 MB RSS and 13.2 s -> 8.7 s, with byte-identical rows. Claude-Session: https://claude.ai/code/session_01BNG66yvb43CB5x3mAWY64h
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
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 fixes the high memory usage regression that was introduced in 2.6.2.
This fixes issue #122
Without this commit, when I try to parse a 33MB xlsx file (500_000 rows) and it consumes 8GB. With this commit it's consuming few hundreds (~300MB)
The reason lies in the namespaces due to all hash instantiation it does for all nodes.
Before this commit:
Screencast from 2025-05-27 09-54-59.webm
With this commit:
Screencast from 2025-05-27 09-58-59.webm