GitLab

erp5_wendelin/DataBucketStream: Migrate len(bucket) to bucket.getBucketCount() to speed up reindexing

When a folder is recursively reindexed, it calculates the length of
child containers [1]. Calling __len__ on a Data Bucket Stream triggers
loading (unghostifying) the entire bucket tree, which can be very slow
for large trees.

However, this cost is unnecessary — the buckets themselves are not meant
to be indexed. To avoid the overhead, this patch refactors len(bucket)
to bucket.getBucketCount(), significantly improving performance.

This change also aligns with the design introduced in nexedi/wendelin@3ef452de,
where buckets were decoupled from the Folder API. Since __len__ on a
Folder reflects the number of folder-contained objects - and buckets are
no longer part of that - returning their count from __len__ would be
misleading.

---

[1] https://lab.nexedi.com/nexedi/erp5/-/blob/1b555dbf/product/ERP5Type/Core/Folder.py#L587-588

/reviewed-on nexedi/wendelin!171
/reviewed-by @Tyagov, @klaus and @paul.graydon