hagen666

OK. I'll have a try!

It's a subset can reproduce the issue: ``` match (n),(m) where id(n)= '46' and id(m)= 'idsx' create (n)-[r:label3]->(m) return r match (n)-->(m) with n, count(*) as p where p >...

When the prefix exists, the c++ target is as fast as the java.

Parse each line not the whole files. here is the C++ code: ``` bool ParseCypher(const std::string& statement) { antlr4::ANTLRInputStream input(statement); CCypherLexer lexer(&input); antlr4::CommonTokenStream tokens(&lexer); tokens.fill(); CCypherParser parser(&tokens); auto stmt =...

May be you need to edit the grammer to support multilie Query in cypher. A runnable example is: ``` oC_Cypher : (SP? oC_Statement ( SP? ';' )? SP? CR? LF?)+...

I only remove 'oC_', and rename 'oC_Cypher' to 'cypher' but i have an old one, which use 4.7.2, the rule name has the first letter in low case, and still...

I'm glad you ran the driver. The parsing time differs obviously in my CentOS 7 and gcc 7.3.0. Some data is here: origin Cypher.g4 file: 240ms The file you provided:...

Yep! It's exciting that you reproduce the performance issue. | Program | Time (M ± SD) |Time w/ warm-up (M ± SD)| | ---- | ---- | ---- | |h+cpp|...

By the way, I want to know how to add the 'warm-up' option.

After the data releases, can you put the link on the repositories' readme or post it to me? Thank you! @GentleZhu