Precision/length parameter of varchar/char types is ignored about arrow-datafusion HOT 6 CLOSED

I updated the issue to make it clear that is not only a problem with literals.

from arrow-datafusion.

Taking into account the fact that DataFusion’s SqlParserOptions already has an option to specify different Sql dialects (and different dialects process this length/precision parameter in CAST expression differently - see examples below), adding a config parameter seems like a very reasonable approach.

SQLLite:

SELECT CAST('12345' AS VARCHAR(2))
12345

BigQuery:

SELECT CAST('12345' AS STRING(2))
Parameterized types are not allowed in CAST expressions.

from arrow-datafusion.

take

from arrow-datafusion.

Maybe we should simply error if DataFusion encouters a type of VARCHAR with an explcit length 🤔 Though if I worry that some people might expect it to work the way it currenty does (ignored)

Perhaps we could make a new configuration setting to control the behavior (error_on_varchar_with_length or something),

from arrow-datafusion.

Maybe we should simply error if DataFusion encouters a type of VARCHAR with an explcit length 🤔 Though if I worry that some people might expect it to work the way it currenty does (ignored)

Perhaps we could make a new configuration setting to control the behavior (error_on_varchar_with_length or something),

Current idea should be adding a new config parameter and when encounter VARCHAR with specific length, we throw an error. Don't know if it works for you @lowka

from arrow-datafusion.

Taking into account the fact that DataFusion’s SqlParserOptions already has an option to specify different Sql dialects (and different dialects process this length/precision parameter in CAST expression differently - see examples below), adding a config parameter seems like a very reasonable approach.

SQLLite:

SELECT CAST('12345' AS VARCHAR(2))
12345

BigQuery:

SELECT CAST('12345' AS STRING(2))
Parameterized types are not allowed in CAST expressions.

Sure, I will do it

from arrow-datafusion.