The data type of predicate in the row filter should be same in the binary expr

[liukun4515]

@alamb @thinkharderdev
I find a issue caused by this rewrite in my work #3396
if we have two parquet, one has c1,c2, the other has c1,c3.
If the filter is c2 = 1, then it will produce a expr NULL = 1 for the parquet with c1,c3 column.
After I remove the binary type coercion in the physical phase, the binary physical expr of NULL = INT(1) can't be created, because we don't need to support type coercion in creation physical expr.

Originally posted by @liukun4515 in #3380 (comment)

[liukun4515]

why not build the build_row_filter from the phy expr?

[liukun4515]

If the table schema is c1,c2,c3, the file schema is c1,c3 with filter c2 = lit.

Why not push the predicate = false to the parquet file directly to replace the c2 = NULL

[alamb]

Why not push the predicate = false to the parquet file directly to replace the c2 = NULL

seems like a reasonable thing to me

[liukun4515]

If the table schema is c1,c2,c3, the file schema is c1,c3 with filter c2 = lit.

Why not push the predicate = false to the parquet file directly to replace the c2 = NULL

cc @thinkharderdev

[alamb]

I think the more general flow here is to take a predicate like c2 = lit and a schema of c1, c3 and replace c2 with NULL -- which is what the code changed in #3470 is.

Then the predicate is NULL = lit which can be reduced to false if we run the constant evaluator on it