Add support for Apache Drill CSV column indices

[davidhaller]

Apache Drill is able to perform queries like this on CSV files:

select columns[0], columns[1] from dfs.`/tmp/drill/test.csv`

JSqlParser could not parse column identifiers with an array index, so I modified the grammar. I'm not sure that I did everything right, as I never worked with JavaCC before. Tests still pass, and the debug output says that columns[0] and columns[1] were recognized as identifiers.

[coveralls]

Coverage decreased (-0.1%) to 84.108% when pulling 9afb380 on davidhaller:master into 66e44c9 on JSQLParser:master.

[wumpz]

The problem I see is, that there is already some kind of array indexing. Additionally the array index will be part of an identifier, which is IMHO not correct.

[davidhaller]

Apache Drill names columns from CSV files columns[0], columns[1] and so on which are usually renamed and casted to the desired data type, like this:

select cast(columns[0] as int) as user_id, columns[1] as username,
cast(columns[2] as timestamp) as registration_date
from `users.csv`

PostgreSQL, for example, has arrays similar like this which can be used within the "select" clause (https://www.postgresql.org/docs/9.6/static/arrays.html). If array support already exists in JSQLParser, what do I need to change to allow array access within "select"?

[manticore-projects]

Obsolete, since it is working fine with JSQLParser 4.7. You can test it online here.