Repository navigation
parseJson implementation in VariantBuilder #3414
Description
Describe the enhancement requested
Every consumer of parquet-variant currently has to implement their own
JSON-to-Variant parsing (Spark, Kafka Connect, Flink, etc.). This adds
native parseJson() static methods to VariantBuilder using Jackson
streaming parser, ported from Apache Spark's production-proven logic.
- Single-pass streaming (no intermediate JsonNode tree)
- Smallest integer type selection (byte/short/int/long)
- Decimal-first float handling, double fallback for scientific notation
- Overloads for String input and pre-positioned JsonParser
Component(s)
Core
Activity
[apacheGH-3414] Add parseJson to VariantBuilder for JSON-to-Variant c…
[apacheGH-3414] Add parseJson to VariantBuilder for JSON-to-Variant c…
steveloughran commentedon Mar 19, 2026
#3451 relates to this; that's an issue/pr for benchmarking variant operations.
[apacheGH-3414] Add parseJson to VariantBuilder for JSON-to-Variant c…
[apacheGH-3414] Add parseJson to VariantBuilder for JSON-to-Variant c…
[apacheGH-3414] Add parseJson to VariantBuilder for JSON-to-Variant c…
[apacheGH-3414] Add parseJson to VariantBuilder for JSON-to-Variant c…
GH-3414: Add parseJson to VariantBuilder for JSON-to-Variant conversi…