You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Jacques Nadeau (JIRA)" <ji...@apache.org> on 2015/04/10 22:48:14 UTC

[jira] [Updated] (DRILL-2682) SQL type syntax "FLOAT(30)" yields AssertionError

     [ https://issues.apache.org/jira/browse/DRILL-2682?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jacques Nadeau updated DRILL-2682:
----------------------------------
    Component/s: SQL Parser

> SQL type syntax "FLOAT(30)" yields AssertionError
> -------------------------------------------------
>
>                 Key: DRILL-2682
>                 URL: https://issues.apache.org/jira/browse/DRILL-2682
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: SQL Parser
>            Reporter: Daniel Barclay (Drill)
>
> Trying to use the syntax "FLOAT(30)" (to request an approximate numeric type (floating-point type) with at least 30 bits of precision (Drill's double-precision floating-point type)) fails with an assertion error:
> > SELECT CAST( 0 AS FLOAT(30) ) FROM INFORMATION_SCHEMA.CATALOGS;
> Query failed: AssertionError: 
> Error: exception while executing query: Failure while executing query. (state=,code=0)
> >
> 1. Drill should not throw an AssertionError (at least not one without some interim "not implemented yet" or "not supported" text) for this.
> 2. If Drill really chooses not to support this, it should give a more direct "not supported" error.  
> (Note that support should be simple:
> - If the expression value is in the range from 25 through 53 (the number of bits of precision in IEEE 754 double precision, Java double, and Drill internal type FLOAT8, then use FLOAT8.
> - Otherwise, if the expression value is in the range from 0 through 24 (the number of bits of precision in IEEE 754 single precision, Java float, and Drill internal type FLOAT4), then use FLOAT4 or FLOAT8.
> - Otherwise, report the appropriate error (the requested precision is greater than implementation-defined maximum precision, the precision is value is negative, or the precision expression couldn't be evaluated).)
> 3.  Drill should support this at least eventually.



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