Claus Ibsen created CAMEL-24869:
-----------------------------------
Summary: camel-jbang - validator hint for sql endpoint queries
with :name parameters (camel-sql uses :#name)
Key: CAMEL-24869
URL: https://issues.apache.org/jira/browse/CAMEL-24869
Project: Camel
Issue Type: Improvement
Components: camel-jbang, camel-yaml-dsl
Reporter: Claus Ibsen
Fix For: 4.23.0
Found in the camel-jbang-mcp stepwise benchmark on the connect-service/sql
example (H2): in 5 of 6 runs the local model's first insert used a parameter
syntax that camel-sql does not have:
{noformat}
VALUES (:customer, :country, 1)
VALUES (:body[customer], :body[country], 1)
{noformat}
camel-sql named parameters are {{:#name}} (looked up in headers, then a Map
body) or {{:#${simple}}}; a bare {{:name}} goes to the JDBC driver as it is and
fails at runtime with a syntax error, after the file consumer has retried it a
hundred times. Nothing in the validator or the write tool says so, and the
model's own knowledge is the Spring/JPA {{:name}} form.
Add a validator row for {{sql}} and {{sql-stored}}/{{jdbc}} endpoint queries: a
{{:name}} that is not {{:#name}}, {{:#${...}}}, {{:?name}} (stored procedure
out parameter) or a {{::}} cast gets the hint "camel-sql named parameters are
:#name (a header or a key of a Map body) or :#${simple}; use :#customer". The
check is deterministic (a regex on the query option), so the write tool refuses
the file with the corrected form before the app runs it.
The dialect mistakes in the same runs (Postgres ON CONFLICT and MySQL ON
DUPLICATE KEY UPDATE against H2) are not a validator matter; the sql tool of
CAMEL-24867 lets a model try a statement against the actual database first.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
