Metasyntax/Macros

[Terry Reedy]

"Paddy McCarthy" <paddy3118 at netscape.net> wrote in message
news:2ae25c6b.0305151348.1ddda895 at posting.google.com...
> "Terry Reedy" <tjreedy at udel.edu> wrote in message
news:<iUudnYfQA-6tHV-jXTWcqw at comcast.com>...
> > "Paddy McCarthy" <paddy3118 at netscape.net> wrote in message
> > news:2ae25c6b.0305140003.dbd812e at posting.google.com...
> > > I went googling for messages on Python and adding a macro
facility,
> > > but ended up with just too many hits and got lost.
> > >
> > > The reason I was looking was that I can get envisage an
application
> > > but seemed to remember that the addition of macros to Python got
> > > shot down before but could not remember the reasons.

Yes.

> > They are all in the hits you gave up on ;-)  (Try just those in
2002.)
> > One quick summary is that Guido wants Python code to look like
Python
> > code and be more or less universally readable by other Python
> > programmers.  He is not going to be the one to help fork Python
into
> > separate sublanguage communities.

My impression is that GvR has opposed non-ASCII/Unicode identifiers,
and certainly non-English 'native language' keywords, both of which
*have* been suggested as additions to standard Python, for much the
same reason.

Please remember that I am acting here as a reporter as much as a
protagonist.  You more are less asked for reasons.

> Ahh, are you saying then that with other languages that have such a
> facility, there is a lot of support for syntax generated by sets of
> macros?

Not sure what you are asking but let me try this.  The C preprocessor
is used for various reasons.  1) Macros enable generic functions like
max(x,y) (x <= y? x: y).  Python functions are generic already.  2)
They enable inline 'functions'.  This improves execution speed but
with some readability problems (and possible bugs) when there are
side-effects.  3) They define constants, which again is about speed,
and thereby less a concern for Python.  4) They can be used to write
'C' code that look a lot like, for instance, Pascal.  I believe there
were once programming shops that had private Pascal/C coding
standards.  But this ghettoized programmers who worked therein.

>From what people have said, Lisp macros are even more powerful, for
both use and abuse, but they are ourside my experience.

> > >  module my_mod (x,y,z) ...
> > >
> > > and create a Python class.
> >
> > This shows the problem already.  'Module' is already in use to
refer
> > to modules, which are different from classes.  If a person has
such a
> > misunderstanding, what will they make of sys.modules?  -- or of
the
> > word 'module' in all current Python
> > writings.

> I guess my example was too simple.

I think it was perfect.  This falls under category 4 above.

...
> Such code may be less readable to the 'out=of-domain' Pythoneer,

Python is about readability.

> but *should* be more readable to the Verilog programmer,

At the price of being ghettoized.  You are free to use domain-specific
names for values, functions, and classes (user-defined types) and call
the result Python.  You can also use (and abuse) the metaclass
facility to do some of the things people do with sophisticated macro
facilities (and call the result Python, though with enough 'abuse',
one might be stretching it).  You can even take the source and rename
the keywords and do whatever, but then maybe you should call it
PMCPython or VeriPython and take responsibility for supporting the
users.  (For all I know, this sort of thing has already been done.)

Once a second set of keywords were allowed, there would soon be a
hundred.  Whereas I like Python for having a set small enough that I
can usually remember them.

Terry J. Reedy