Jackson 3 port - #95
Open
vincentjames501 wants to merge 3 commits into
Open
Conversation
vincentjames501
force-pushed
the
jackson-3-port
branch
from
August 3, 2026 22:59
3b804de to
34a4698
Compare
Three decoder optimizations, measured together as +15-23% decode throughput with string keys and +28-48% with keyword keys vs 1.0.0 at 100B+ payloads (bench/2026-08-01-full-comparison.md): - KeywordKeyDeserializer fronts Keyword.intern (fresh Symbol + global weak-ref table per call) with a bounded String -> Keyword cache using the same clear-when-full strategy as jackson-core's own tools.jackson.core.util.InternCache. - JSON objects of 8 entries or fewer decode to PersistentArrayMaps, matching small Clojure map literals (and cheshire), instead of paying hashing + HAMT node churn in a transient PersistentHashMap; larger objects spill to the transient as before, and duplicate keys keep assoc semantics via createAsIfByAssoc. Property loops use nextName(), jackson-core's fused nextToken()+currentName() fast path. - ClojureUntypedDeserializer makes untyped (Object-targeted) decoding a single recursive switch instead of double-dispatching through databind's UntypedObjectDeserializer. Non-stock Map/List deserializers (java-collection-module, jsonista.tagged, user modules) are still delegated to; the wiring lives in resolve() because runtime findNonContextualValueDeserializer lookups skip contextualization. Also tried and rejected on measurement: SerializedString caching in KeywordSerializer (2-4% regression on the write-value-as-string path; bench/2026-07-31-kw-optimizations.md). Its escaping-behavior tests are kept.
Benchmark harness: - :encode-kw/:decode-kw/:encode-kw-bytes groups measure the keyword-keys-object-mapper workload against data.json (:key-fn) and cheshire (parse-string true) - encode-jsonista/decode-jsonista no longer shadowed by tagged-module variants, so :encode/:decode measure stock jsonista Results (bench/2026-08-01-full-jackson321.txt on Jackson 3.2.1, method + controls in bench/2026-08-01-full-comparison.md): vs jsonista 1.0.0/Jackson 2, decode is +15-23% with string keys and +28-48% with keyword keys at 100B+; encode is unchanged at 1KB+, and the sub-100B encode regression is in the raw Jackson 3 engine, not jsonista. README throughput tables and graphs regenerated from this run; CHANGELOG carries the summary.
vincentjames501
force-pushed
the
jackson-3-port
branch
from
August 3, 2026 23:06
34a4698 to
4a8eb98
Compare
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.
See #89
This isn't intended to be merged as-is but rather start a discussion.
#89 (comment)
In the spirit of full transparency, this was mostly done by Claude Fable (probably obvious from some the word slop). If we are interested in going forward, I'm happy to do more human iteration on this and clean things up a bit.