On 5/31/2018 10:49 AM, Dan Strohl via Python-list wrote:
Is it possible to override the assignment of built in types to the shorthand
representations?
By which I presume you mean literals and overt (non-comprehension)
displays. So you wish that Python should be even more dynamic. (Some
wish it were less so ;-)
The CPython parser is generated by a parser-generator program from the
python grammar and a table of non-default functions to call when the
parser recognizes certain grammatical productions. For instance,
*stringliteral* is mapped to str. I assume that the mapping is
currently direct, and not routed through the builtins dict. I don't
know what other implementations do.
To change this, I believe you would have to introduce indirect mapping
functions. One possibility would be C equivalents of functions like
def get_str(): return builtins['str']
Then you could change future parsing by executing
builtins['str'] = mystr\
However, this would affect the parsing of imported modules, if and when
they are parsed, and exec and eval calls made within imported functions.
This would not be good.
So I believe you would also need to introduce a module copy of a subset
of builtins, call it '__modclass__', whose keys would be the classes
called by the parser, and use that in the get_xyz functions. Then I
believe you could change parsing within a module by executing
__modclass__['str'] = mystr
and yes, I know I could simply not do [] and always do my_list('item1',
'item2', 'item3')
I think we should stick with this.
This would only be scoped to the current module and would not be imported when
that module was imported.
The harder part, I think, is "and not affect parsing of imported modules
if they are not already parsed and not affect exec and eval calls in
imported modules.
--
Terry Jan Reedy
--
https://mail.python.org/mailman/listinfo/python-list
