RowToStructByName Snake Case Collision

[nolandseigler]

rows.go fieldPosByName currently removes all "_" characters. This causes structs with "db" tags like "account_id" and "account__id" to be treated as the same tag. The index returned by fieldPosByName is always the first field tagged which causes an error like "struct doesn't have corresponding row field account__id".

I have written a test that demonstrates the issue and implemented some proof of concept code to demonstrate a potential fix. All tests pass.

This is my first PR. Please let me know if I am now following the proper procedure. I can certainly open up a discussion or issue as well. Thank you.

[jackc]

I suppose this is more correct behavior when a db tag is provided. Though it's hard to see how having a database table where the columns names only differed by underscores is a good idea...

That said, if this change is going to be made it should cover all normalization. That is, if a db tag is provided all name normalization is disabled. I would suggest also doing a simple == comparison instead of the case insensitive compare.

[nolandseigler]

I suppose this is more correct behavior when a db tag is provided. Though it's hard to see how having a database table where the columns names only differed by underscores is a good idea...

That said, if this change is going to be made it should cover all normalization. That is, if a db tag is provided all name normalization is disabled. I would suggest also doing a simple == comparison instead of the case insensitive compare.

Great idea. Thanks for the quick response. I will do that right now.