From a9204f341da230981e0b589566c204159a3c3dde Mon Sep 17 00:00:00 2001 From: abram axel booth Date: Fri, 7 Feb 2025 13:00:57 -0500 Subject: [PATCH] add _TODO_denorm.txt (temp) --- _TODO_denorm.txt | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) create mode 100644 _TODO_denorm.txt diff --git a/_TODO_denorm.txt b/_TODO_denorm.txt new file mode 100644 index 000000000..7bf77b378 --- /dev/null +++ b/_TODO_denorm.txt @@ -0,0 +1,20 @@ + +ok so + +trovesearch_denorm has perhaps tooo many indexed fields +(especially text fields and especially longer property paths) + +## possible mitigations: + +DONE: +- skip indexing nested "identifier" text (saves 107 fields) +- allow wild-card properties only to depth 1 (disallow `cardSearchText[*.*]`) +- stop hashing so much rendering valuesearch responses (get unique ids other ways) + +could still do: +- skip making iri-value docs for common values? + - are there common values that will definitely not be value-searched? +- revisit text indexing/querying + - less fuzz? less slop? different analyzers? +- shorter dynamic field names -- hash? make reversible with additional sub-index +- more splitting! split iri-values index (maybe by hash-prefix? or card resource type?)