[Python-3000] Release Countdown

[Jim Jewett]

On 8/31/07, "Martin v. Löwis" <martin at v.loewis.de> wrote:
> >> Yuck, yuck about the source file encoding part. Also, there is no way
> >> to tell that a particular argument was passed a literal.

> > There is when compiling to bytecode; it goes in co_consts.

> >> The very
> >> definition of "this was a literal" is iffy -- is x a literal when
> >> passed to f below?

> >>   x = "abc"
> >>   f(x)

> > No, it isn't.

> By that definition, bytes never receives a constant.

To go back to the original motivation

    x.split(":")    # a constant, currently fails in Py3K

    x.split(b":")  # mechanical replacement for x.split(":")

    sep=":"
    x.split(sep)  # annoying but less important failure


I would prefer that x.split(":") work.

If that happens because bytes.split does the conversion for me (so
that x.split(sep) also works), then great.  But I realize that would
require an assumption about the proper encoding.

If it works because the bytecode compiler changes x.split(":") into
the moral equivalent of

    try:
        x.split(":")
    except StrNotBytesError:
        x.split(b":")

that is good enough.  And for constants which appear as string
literals in the code (token stringliteral), the proper encoding is
known.

-jJ