Stephen Mallette created TINKERPOP-3309:
-------------------------------------------
Summary: Absent parameter key is treated as a null value for
TinkerGQL
Key: TINKERPOP-3309
URL: https://issues.apache.org/jira/browse/TINKERPOP-3309
Project: TinkerPop
Issue Type: Improvement
Components: gql
Affects Versions: 4.0.0
Reporter: Stephen Mallette
The "Parameterized Queries" table in docs/src/reference/the-traversal.asciidoc
lists two distinct behaviors for match(String, Map): a parameter key present
with a null value "passes only for elements where the property is absent",
while a key absent from the map "always fails — no element matches". In
practice the two cases behave identically — an absent key passes for elements
where the property is absent, and only matches nothing when every candidate
happens to have the property.
PropertyPredicate.test resolves the parameter with params.get(name), which
returns null for an absent key just as it does for a key explicitly mapped to
null, so there is no "always fails" path. DefaultGqlExecutor.seedIterator
resolves seed-vertex parameters the same way. Observed on master at d45c55bf
and still present at 85c4269d.
Either the table row is wrong, or the implementation is. An absent key is
usually a caller mistake such as a misspelled parameter name, and treating it
as null silently widens the result set instead of failing loudly;
params.containsKey(name) would distinguish the two cases. A documentation fix
is ready, but it shouldn't land until the intended semantics are decided.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
