To avoid maintain sqlflow.proto file in the separated repo

[Yancey0623]

Mabe adds the sqlflow repo as the submodule of pysqlflow is a good way.

[tonyyang-svail]

I believe it is better to have a little duplication than importing a huge library for a single file. :) ref

[typhoonzero]

Just for discussion: #70 (comment)

[LiuGuodong1314]

How can I run extended SQL using pysqlflow

[Yancey0623]

Hi @LiuGuodong1314 , you can use the all-in-one Docker image, please go to https://github.com/sql-machine-learning/sqlflow/blob/develop/doc/demo.md

[LiuGuodong1314]

Hi @Yancey1989,
I execute
"SELECT * FROM iris.train TRAIN DNNClassifier WITH n_classes = 3 COLUMN sepal_length, sepal_width, petal_length, petal_width LABEL class INTO sqlflow_models.my_dnn_model"
in python language using pysqlflow, but it throws "runExec failed: Error 1064: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version ...."
How can I manage it?

[Yancey0623]

It seems that the extended SQL should be ... WITH n_classes = 3, hidden_units = [10, 20] COLUMN ... ? And your problem is not related to this issue, maybe you can recreate an issue at https://github.com/sql-machine-learning/sqlflow/issues if the problem persists.