sql/mysql: Attempt a fix for RawExpr cases

[ofpiyush]

This is an optimistic attempt at a fix for expressions/functions in schema. Feel free to correct and guide me towards a better solve. :)

The wrapping is somewhat redundant when we already have sql() but going by the comments in code, backwards compatibility has made the situation a little messy.

Fixes #795

Note: Initially, MySQL looked like it was working fine without this change.
But if we tried to inspect -> drop database -> create. We ended up with string default instead of an expression.

[rotemtam]

hey @ofpiyush
thanks for working on this! notice that CI is red, do you mind taking a look at why?
🙏

[ofpiyush]

Yes, I was playing with atlas and the bug just sucked me in. That commit was a lesson in why you shouldn't push code at 3 AM 😅

CURRENT_TIMESTAMP cases were failing. For now, I've added an exception for them.

A better solve for this family of problems will be fairly involved. Comparing the exact query can be brittle but attempting to compare the intent(via AST) behind the query might still result in wrong query and/or lot of edge cases to handle on a case to case basis.

Another such example is a field with default CONCAT(my_field,' anystr'). This generates an escaped query (tried with MySQL 8), which again breaks in the inspect -> drop database -> apply scenario.

[a8m]

Thanks for the contribution @ofpiyush 🙏

Code looks great, but the tests in integration/mysql_test.go need to be updated with the new inspection format. Let me know if you need any help with this.

Another such example is a field with default CONCAT(my_field,' anystr'). This generates an escaped query (tried with MySQL 8), which again breaks in the inspect -> drop database -> apply scenario.

Can you elaborate with an example? Did you address it in this PR? If not, that's fine, I can take this as well.

[ofpiyush]

@a8m Would you believe that those tests pass in go 1.17? Would be interesting to know why!

I've updated the test cases.

Can you elaborate with an example? Did you address it in this PR? If not, that's fine, I can take this as well.

I've opened #803 to describe the bug. I haven't attempted it mostly because I am not sure if we're missing an unescape or is it because unescape itself isn't doing what it is supposed to.

[a8m]

Thanks for the contribution @ofpiyush 🙏