Unlike Java, whose strings are always in the UNICODE character set, SQL's strings can be in a variety of character sets. When you perform the '=' operation, both arguments need to be the same type. The character literal 'bluejoe' implicitly has character set "ISO-8859-1" and collation "ISO-8859-1$en_US$primary", and I guess the NAME column has something else, so Calcite adds a cast to allow them to be compared.
On Tue, Sep 6, 2016 at 8:30 PM, bluejoe <[email protected]> wrote: > Hi, all > I wrote a custom ScannableTable and override the scan(context, filters) > method to implement the searching function. > When I sent a SQL command like > > select * from PERSONS where NAME=‘bluejoe’ and AGE>30 > > In the filters parameter, I found the filter on column `AGE` is parsed > properly like `>($2, 30)` > > While the filter on column `NAME` is parsed as: > > =(CAST($1):CHAR(7) CHARACTER SET "ISO-8859-1" COLLATE > "ISO-8859-1$en_US$primary" NOT NULL, 'bluejoe’) > > Why the column identifier`NAME` is transformed? How can I get the real column > $1? Or is there anyway to prevent the transformation? > > Many thanks! > > Best regards, > Zhihong SHEN > ——————————————— > Zhihong SHEN, Ph. D., Senior Engineer > Big Data Application Service Technology Laboratory, > Computer Network Information Center, Chinese Academy of Sciences > office phone:+86-10-58812516 > mobile：+86-13671116520 >
