Change Details

While BagOStuff supports the storing of large values (>1mb) in memcached using WRITE_ALLOW_SEGMENTS, the WANObjectCache class does not support it yet. Since the preprocessor text was changed to use WANCache, large wiki pages will not benefit from caching since the cache writes will fail. The preprocessor cache is useful for expanding templates during parser cache misses (e.g. on edit or for users with non-standard preferences). An example of large parser output entries was found during unrelated debug log entries for RefreshLinksJob failures at T307133. We should: * [x] Add a "segmentable" option to WANCache that uses WRITE_ALLOW_SEGMENTS * [] Support segmentation of non-text values or make PreprocessorHash use JSON * [] Use the new option in PreprocessorHash::preprocessToObj()

While BagOStuff supports the storing of large values (>1mb) in memcached using WRITE_ALLOW_SEGMENTS, the WANObjectCache class does not support it yet. Since the preprocessor text was changed to use WANCache, large wiki pages will not benefit from caching since the cache writes will fail. The preprocessor cache is useful for expanding templates during parser cache misses (e.g. on edit or for users with non-standard preferences). An example of large parser output entries was found during unrelated debug log entries for RefreshLinksJob failures at T307133. We should: * [x] Add a "segmentable" option to WANCache that uses WRITE_ALLOW_SEGMENTS * [] Use the new "segmentable" option in SqlBlobStore * [] Support segmentation of non-text values or make PreprocessorHash use JSON * [] Use the new "segmentable" option in PreprocessorHash

While BagOStuff supports the storing of large values (>1mb) in memcached using WRITE_ALLOW_SEGMENTS, the WANObjectCache class does not support it yet. Since the preprocessor text was changed to use WANCache, large wiki pages will not benefit from caching since the cache writes will fail. The preprocessor cache is useful for expanding templates during parser cache misses (e.g. on edit or for users with non-standard preferences). An example of large parser output entries was found during unrelated debug log entries for RefreshLinksJob failures at T307133. We should: * [x] Add a "segmentable" option to WANCache that uses WRITE_ALLOW_SEGMENTS * [] Use the new "segmentable" option in SqlBlobStore * [] Support segmentation of non-text values or make PreprocessorHash use JSON * [] Use the new option in PreprocessorHash::p"segmentable" option in PreprocessToObj()orHash