Description
As of r925353 (but also in 0.5.1) it seems that the AST is not complete
For example in 0.5.1
cassandra> get Messages_test.Messages['twitter.com:10731838401
line 1:51 mismatched character '<EOF>' expecting '''
line 0:-1 mismatched input '<EOF>' expecting StringLiteral
Exception in thread "main" java.lang.AssertionError
at org.apache.cassandra.cli.CliClient.executeGet(CliClient.java:279)
at org.apache.cassandra.cli.CliClient.executeCLIStmt(CliClient.java:57)
at org.apache.cassandra.cli.CliMain.processCLIStmt(CliMain.java:131)
at org.apache.cassandra.cli.CliMain.main(CliMain.java:172)
A little debug added in trunk shows that ANTLR is emitting "mismatched token" nodes:
cassandra> get Messages_test.Conversations[f
line 1:32 mismatched input 'f' expecting StringLiteral
(NODE_THRIFT_GET <mismatched token: [@-1,0:0='<no text>',<-1>,0:-1], resync=Messages_test.Conversations[f>)
0
Exception in thread "main" java.lang.AssertionError
at org.apache.cassandra.cli.CliClient.executeGet(CliClient.java:312)
at org.apache.cassandra.cli.CliClient.executeCLIStmt(CliClient.java:60)
at org.apache.cassandra.cli.CliMain.processCLIStmt(CliMain.java:213)
at org.apache.cassandra.cli.CliMain.main(CliMain.java:270)
thus the assertion is correct in complaining that it did not get a NODE_COLUMN_ACCESS but it should probably be handled instead of crashing.
but sadly, I can't really understand where this happens, as I'm no expert of antlr, but I wonder if it could be just instructed to fail fast without reaching the ast walking?