You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Julian Hyde (JIRA)" <ji...@apache.org> on 2015/04/26 23:32:38 UTC

[jira] [Commented] (DRILL-2880) non-syntax error has unexpected category "PARSE ERROR"

    [ https://issues.apache.org/jira/browse/DRILL-2880?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14513277#comment-14513277 ] 

Julian Hyde commented on DRILL-2880:
------------------------------------

Might be a sign that the parser is trying to do too much validation. IMHO the parser should return a complete AST before [semantic] validation starts. That means the AST needs to be able to refer to a file that does not exist, and indeed to numbers which are too large to be held in BIGINTs.

> non-syntax error has unexpected category "PARSE ERROR"
> ------------------------------------------------------
>
>                 Key: DRILL-2880
>                 URL: https://issues.apache.org/jira/browse/DRILL-2880
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: SQL Parser
>            Reporter: Daniel Barclay (Drill)
>            Assignee: Aman Sinha
>            Priority: Minor
>
> Some errors that are not syntax errors (failures to match the productions of the SQL grammar), such as referring to a non-existent table:
> SELECT * FROM NoSuchFile;
> yield error messages that unexpectedly refer to parsing, such as:
> PARSE ERROR: From line 1, column 15 to line 1, column 26: Table 'NoSuchFile' not found
> &nbsp;
> That "PARSE ERROR" sounds like it's saying it's specifically a parsing error, when it's really an error downstream of parsing.  
> (A parsing error requires changing the query to conform to the grammar.  In the above case, creating a file or view would resolve the problem without changing the query, so the problem is not a parsing error.)
> &nbsp;
> Its not clear whether that error should be reported using a different existing or new error category, or whether the category "PARSE ERROR" should be renamed (to cover both real parsing errors and the level of errors).
> Note, however, that to avoid confusion, errors that can be fixed by changing something other than the query (i.e., the data store state) probably should not be reported using the same category as true syntax/parsing errors.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)