StatementSwitchToExpressionSwitch: skip rule cases #3656
Conversation
| @@ -122,7 +134,7 @@ private static AnalysisResult analyzeSwitchTree(SwitchTree switchTree) { | |||
|
|
|||
| List<? extends StatementTree> statements = caseTree.getStatements(); | |||
| CaseFallThru caseFallThru = CaseFallThru.MAYBE_FALLS_THRU; | |||
| if (statements == null || statements.isEmpty()) { | |||
| if (statements.isEmpty()) { | |||
There was a problem hiding this comment.
This reverts the change in #3645, because there are actually many more CaseTree#getStatements() calls in this class, and they all assume a non-null result. The isRuleSwitch check now guarantees that.
| private static boolean isRuleSwitch(SwitchTree switchTree) { | ||
| return switchTree.getCases().stream().anyMatch(StatementSwitchToExpressionSwitch::isRuleCase); | ||
| } | ||
|
|
||
| private static boolean isRuleCase(CaseTree caseTree) { | ||
| try { | ||
| return "RULE".equals(CaseTree.class.getMethod("getCaseKind").invoke(caseTree).toString()); | ||
| } catch (ReflectiveOperationException e) { | ||
| throw new LinkageError(e.getMessage(), e); | ||
| } | ||
| } |
There was a problem hiding this comment.
Given the documented invariant on which we now rely, perhaps it's more accurate to test for "STATEMENT". But it seems unlikely that a third Kind will be introduced any time soon, and this way we avoid a negation at the call site. But I'm happy to switch this around.
| @@ -1030,6 +1030,25 @@ public void emptySwitchCases_noMatch() { | |||
| " }", | |||
| " }", | |||
| "}") | |||
| .setArgs("-XepOpt:StatementSwitchToExpressionSwitch:EnableDirectConversion") | |||
There was a problem hiding this comment.
Added this for symmetry with other tests, and because in theory it covers more code paths (except that in practice isRuleSwitch(switchTree) is tested before the flag is consulted).
NB: the other tests use .setArgs(ImmutableList.of(...)); I wouldn't mind opening a separate cleanup PR for that.
The cases in a switch statement must all be of kind `STATEMENT` or `RULE`. In the latter case `StatementSwitchToExpressionSwitch` does not apply.
|
Rebased and resolved conflict. One way to look at this PR is that without it, it is unlikely that JDK 14+ users can enable |
Stephan202
force-pushed
the
bugfix/dont-migrate-rule-kind-switches
branch
from
4a7b6bf
to
b0632d0
Compare
|
The solution in #3676 doesn't require reflection; will close this PR. |
The cases in a switch statement must all be of kind
STATEMENTorRULE. In the latter caseStatementSwitchToExpressionSwitchdoes notapply.