spark-rapids
spark-rapids copied to clipboard
[FEA] Support parse_url
I wish we can support parse_url function.
eg:
spark-sql> select parse_url(c_customer_id,'HOST') from tpcds.customer limit 10;
! <ParseUrl> parse_url(c_customer_id#1, HOST, false) cannot run on GPU because GPU does not currently support the operator class org.apache.spark.sql.catalyst.expressions.ParseUrl
Discuss with the python cudf team whether they could benefit from a kernel for this functionality in libcudf.
Is it sufficient for the initial implementation that the 2nd and 3rd parameters are literals?
From the logs, i think it is good enough. I am also double checking with user.
The requirement is to support all 3 input parameter for this function including the KEY.
We need to support PATH, QUERY and HOST.
Parse_url is now supported for Host with this PR.