Claus Ibsen created CAMEL-23596:
-----------------------------------
Summary: camel-yaml-io - Replace hand-written YamlWriter with
generated direct YAML writer
Key: CAMEL-23596
URL: https://issues.apache.org/jira/browse/CAMEL-23596
Project: Camel
Issue Type: Improvement
Reporter: Claus Ibsen
Fix For: 4.21.0
The YAML dump mechanism in camel-yaml-io uses a hand-written YamlWriter (600
lines) with 43+ hardcoded element names and special cases. Every new EIP that
has configuration children (like circuitBreaker + resilience4jConfiguration) or
introduces a new container type (like rests) silently produces wrong YAML until
someone manually adds another special case.
The root cause: the generated ModelWriter emits a flat
startElement/attribute/endElement stream (designed for XML), and YamlWriter
must re-classify every child element at runtime to decide if it is a step, a
configuration property, an expression, or a structural child. This
classification is entirely hardcoded.
Proposal: Replace the hand-written YamlWriter + EipNode + Jackson pipeline with
a generated direct writer that builds JsonObject/JsonArray structures and a
small hand-written YamlPrinter that serializes them to YAML text. No Jackson
dependency needed.
Current pipeline (6 conversions):
Model -> Generated ModelWriter -> YamlWriter.startElement/endElement ->
EipModel metadata -> EipNode -> JsonObject -> JSON String -> Jackson
ObjectMapper -> JsonNode -> Jackson YAMLMapper -> YAML String -> post-process
(strip indent)
New pipeline (2 conversions):
Model -> Generated YamlModelWriter -> JsonObject/JsonArray -> YamlPrinter ->
YAML String
Benefits:
- No more hardcoded special cases that break when new EIPs are added
- Classification of config children vs steps vs expressions happens at
code-generation time
- Removes jackson-dataformat-yaml dependency from camel-yaml-io
- Simpler pipeline with full control over YAML output format
- Automatically correct when the model changes (new options, new EIPs)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
