[Python-3000] bytes and dicts (was: PEP 3137: ImmutableBytesand Mutable Buffer)

[Terry Reedy]

"Guido van Rossum" <guido at python.org> wrote in message 
news:ca471dc20709290733i54f63ac3pb4501b94530db820 at mail.gmail.com...
| Until just before 3.0a1, they were unequal.

I think it valuable that in the language as delivered, 'o==p' (as well as 
'bool(o)' )always return True or False.  Both make reasoning about code 
easier since one does not have to learn and carry around in the back of 
one's mind niggling exceptions.  I am -1 on the last minute change and for 
much the same reasons I have against building into the language 
Windows-specific suppression of \r output (see pydev post).

|  We decided to raise
| TypeError because we noticed many bugs in code that was doing things
| like
|
|  data = f.read(4096)
|  if data == "": break
|
| where data was bytes and thus the break never taken.

As G. Smith said, if a generic comparison is meant, then that should be

if not data: break

In any case, this seems like a old-code translation problem rather than a 
new-code writing problem.  We already know that each existing str literal 
may have to be humanly checked to determine whether a 'b' should be 
prepended, as would appear to be the case above.

| Similar with checks for certain magic strings (so it wasn't just empty 
strings).

If a generic comparison is wanted, then "if data in ('abc', b'abc')".

If a specific comparison is wanted, then raising an exception complicates 
what should be simple.  Consider

def g(stuff):
  if stuff == 'abc": special_text()
  elif stuff == b'abc': special_bytes()
  else: general_stuff(stuff)

Breaking equality is not free.

| It is also in line with the policy to refuse things like
| b"abc".replace("a", "A") or "abc".replace(b"b", b"B").

I do not see the connection.  I would expect either to return TypeError, 
just as
  '123'.replace(1,4)
does today, even though
  '1' == 1
is False, rather than exception raising.

Terry Jan Reedy