1.3.8: unicode_literal always adding the _binary prefix

[claudep]

After the commit 2617620, unicode_literal is always adding the _binary prefix because db.literal inside unicode_literal is called on the encoded value (and encoded values are always bytes).
This is a serious regression!

[methane]

Sorry, but could you describe more detail?

Is this problem is on Python 3 only? or Python 2 only? both?

Original report (#106 (comment) ) was on Python 3.5.

because db.literal inside unicode_literal is called on the encoded value (and encoded values are always bytes).

What the "encoded value" exactly means?

[claudep]

My debug session was on Python 3, but it might happen on Python 2 too.
See this line (inside def _get_unicode_literal()):

return db.literal(str(u).encode(unicode_literal.charset))

Before the patch, db.literal was using string_literal even for bytes, now it uses bytes_literal, so as str(u).encode(unicode_literal.charset) is always bytes, the return value is always prepended with _binary, while the original value is unicode.

[claudep]

The solution might be to replace

return db.literal(str(u).encode(unicode_literal.charset))

by

return db.string_literal(str(u).encode(unicode_literal.charset))

And probably the same for Python 2.

[vtermanis]

@claudep: I think that is the right solution and is an oversight I made when suggesting the original change.