Ideas for Python 3

[David MacQuigg]

On Wed, 12 May 2004 15:39:41 -0700, Josiah Carlson <jcarlson at uci.edu>
wrote:

>>>I've read the link you posted, and I've previously taken a look at 
>>>Prothon.  Prototype syntax does not suit my taste in programming 
>>>languages.  If Python 3.0 becomes Prothon, I'll fork the most recent 
>>>Python 2.x codebase that I prefer, maintaining it myself if necessary.
>>>
>>>Just as my anti-prototype perspective colors my opinions on what is 
>>>reasonable, I would imagine that your pro-prototype perspective colors 
>>>yours.
>> 
>> I am currently neutral on the idea of prototypes, still waiting for
>> anyone to show me a good use case, so I can include it in my webpage.
>> I still have "cloning" of instances in my proposal, but you will
>> notice they are at the bottom of the list of benefits at
>> http://ece.arizona.edu/~edatools/Python/PrototypeSyntax.htm
>
>Considering that you provide an animal heirarchy use-case to illustrate 
>what you propose, I'd say that you're a bit on the positive side of neutral.

The animal hierarchy is intended as a vehicle to demonstrate *all* the
features of the proposal.  The "with" statement, and the ability to
"clone" instances are actually outside of that hierarchy, and are
easily separable.  Taking them out won't hurt the main part of the
proposal, and I probably should do that on the next revision, just to
avoid having the whole proposal valued on a feature that can't be
justified.  On the other hand, maybe I should leave these in as
"throw-aways", so if the proposal ever does get near implementation,
they can cut these features instead of something important. :>)

>I don't believe I understand the idea of "cloning", at least in terms of 
>prototypes.  In my mind it brings up either using a copy() method on 
>instances, or using instance1.__dict__.update(instance2.__dict__) to 
>transfer instance variables.  Unless there is more (or an actual) 
>description about what a "clone" is, I can't say whether it is a benefit 
>or drawback.

It's real hard to get a clear statement from the prototype advocates,
but my understanding is that they want to eliminate all classes, and
just make instances, one from another by "cloning".  Cloning is
different than instantiation in Python, because all attributes are
copied to the clone.   This means you cannot change an ancestor's
attribute and have it change for all descendents that don't explicitly
over-ride that attribute.  This lack of an active "inheritance"
mechanism seems like a major disadvantage to me.  Presumably there is
some counterbalancing advantage, but I have not been able to figure
that out.

>> I typically get enthusiastic about something when I first see it, then
>> I learn a little more, and some of the new tricks/features drop to the
>> bottom of my list.  What remains of my original enthusiasm for Prothon
>> is still in the unification of methods and functions.
>
>I guess I must be an "Expert", because unifying methods and functions 
>doesn't seem to be necessary to me.  While some (including yourself) 
>state that it would make learning portions of Python easier, never 
>having had issues handling lambdas, functions, class methods, static 
>methods, or instance methods, I don't know what people have issues with.
>
>Really the question is: "Is there a problem with current function, 
>method, etc., syntax?"  If there is a problem, the next question is, 
>"What are people having problems understanding?"  At some point there 
>would likely be examples provided that show where the user was confused. 
>  Documenting those examples to explain the syntax, I believe, would be 
>sufficient, but then again, I RTFM.

I've never been able to convince anyone based on an isolated example.
The problem is that even if I show something with a benefit as obvious
as the elimination of special syntax for a static method, the counter
is either static methods are simple, or you don't need them.

Take a look at the 96 pages devoted to OOP in Learning Python, 2nd ed.
This is generally considered the best introductory text on Python, and
it looks to me like it includes the right choice of topics, and is
well written.  Then take a look at
http://ece.arizona.edu/~edatools/Python/Prototypes.doc  In 8 1/2
pages, I provide a basic introduction to everything students will need
to write class hierarchies, work with bound and unbound functions, and
write programs that in Python would require static methods.  If I were
to add some additional topics, examples and exercises, I think this
presentation would expand to 40 or 50 pages.  This is something I
could usefully teach in 4 hours of the student's time ( lecture,
reading, and practice ).

The key to making this work is to build on of what the students
already know ( functions, modules, and global variables ).  With the
unification I am proposing, there is almost nothing to learn in going
from modules to classes.  It's just a different way to package data
and functions.

I could probably do almost the same in Python by ignoring all but the
most common method form.  The problem is that ignoring these other
forms is not the same as ignoring metaclasses.  The student is
unlikely to encounter a need for metaclasses, but the need for static
methods arises naturally.  At least it seems natural to me, having not
been conditioned to avoid them.  Not being able to call function from
a class without a special first argument, seems to me like not being
able to have the letter 'q' in a variable name.

There is extensive discussion of this topic on the thread "Unification
of Methods and Functions", including some examples of how to write
classes which avoid the need for static methods.  So far, I have not
seen anything that is better as an introduction to OOP than my
original example.  You might want to join in that discussion, if you
can add something to our understanding.

>In terms of instance variables and accessing via .variable, I would 
>imagine the lack of 'this' (for C++ or Java programmers) or 'self' (for 
>Python programmers) to be somewhat disconcerting, and though '.' is used 
>to denote a namespace heirarchy in functions, modules, classes, etc., 
>having it lead a an instance variable or function suffers from the same 
>"punctuation-itis" that is listed as a drawback for various function 
>definition syntaxes.

The issue is not .var vs self.var vs $var or any other variant.  They
are all equally explicit, and everyone will have a different opinion
as to which is ugly or pretty.  The issue is the different argument
sequences.  To unify all forms of functions and methods, we need to
get 'self' out of the argument list.  I know this seems like a trivial
change, but take a look at how many pages in Learning Python are
devoted to the different function and method forms.

>For those that I come across who are hung up on the handful of syntaxes 
>available for functions/methods, I point out that really only two are 
>generally needed (function(), instance.method()).  The others are merely 
>there to make life easier for 'advanced' users, and if are not labeled 
>as such, should be (class.method(instance) is pretty useful, as can be 
>lambda, and are the two 'advanced' function types that I use).  By only 
>introducing the two major syntaxes, whose only difference is the naming 
>of the referenced instance (for flexibility), I believe the vast 
>majority of people who had troubles before, would have very few now.

If you can show me that static methods are not necessary, please join
the discussion under "Unification of Methods and Functions".  We have
a fairly simple problem, an animal hierarchy, and some fairly simple
requirements.  These are stated as an exercise at
http://ece.arizona.edu/~edatools/Python/Exercises/

>> The key to finding what is fundamentally good in these other
>> languages, and what is just someone's personal preference for being
>> different, is keeping an open mind.  It's astonishing how few people
>> can do that in comparing computer languages.  I guess it is just
>> easier to be "anti" or "pro" and skip the investigation and thinking.
>
>On that note, I am relatively closed-minded about (even suggested) 
>alterations to Python because it works so well for me /now/, I'm being 
>paid to write software with it /now/, of the alterations that are being 
>offered, I don't believe any would increase my productivity /soon if 
>ever/, and I find them to be both aesthetically unpleasing, as well as 
>relying on the assumption that people don't /currently/ understand Python.

What I am proposing will be of no benefit to advanced users.  But it
won't be a problem for them either.  I would expect an advanced user
to learn the new simpler syntax in about five minutes.  As for the
aesthetics, I would trust GvR to make a good choice, and after using
the new syntax for a while it will look just right.

What I like about GvR is that in spite of his genius, he can still
show concern for something as simple as the problem beginners have
with dividing two integers.  That problem doesn't bother me or anyone
I know, but I like the fact that he fixed it.

>I believe that the current learning curve of Python is sufficiently 
>gradual as to allow the vast majority of people to use it effectively. 
>Knowing the ins and outs of the entirety of the language is not 
>necessary, and even a reasonable subset excludes staticmethods, 
>metaclasses, etc.

Take a look at the link above for a simple problem that uses static
methods.  The lack of static methods would seriously cramp my style.

>>>I would also expect that someone who enjoys using lisp or 
>>>perhaps SML would have their opinion on what is reasonable, colored by 
>>>their preferences.  Considering that Python is not a prototype-based 
>>>language today in 2.3 (or the forthcoming 2.4 in the fall), having it 
>>>become one in Python 3 would be quite a drastic change, quite literally 
>>>like going from Python 2.3 to Prothon.
>> 
>> I continue to find new surprises in Python.  The ability to change
>> classes into prototypes by using descriptors is the latest exammple.
>> This seems like a drastic change, but it is all within the capability
>> of Python.
>
>Being able to use prototypes, without prototype syntax, is not changing 
>the language syntax.  The fact that Python doesn't need prototype syntax 
>to have prototypes, I believe, is an argument against altering Python's 
>syntax.

It's an argument agains making prototypes ( 'with' statements and
cloning ) part of the core.  It's irrelevant to the more important
parts of the proposal.

>> Prothon does have some changes that are more drastic, but in my
>> opinion have no real benefit over Python.  I measure drastic by how
>> much effort it will take to translate existing Python programs to
>> Prothon.  By this measure, the unification of functions and methods is
>> not a drastic change.  I believe it will be possible to automatically
>> translate all Python methods into the new form.  So if we categorize
>> syntax changes as ( Compatible / Migratable / Totally Different ), the
>> proposed changes are in the middle.
>
>Function/method unification is certainly migratable, I've never claimed 
>otherwise.  In fact, with a good refactoring tool, it is trivial (and 
>actually a search+replace will go a long ways towards migrating).  The 
>'.' prefix "punctuation-itis" I described earlier, coupled with your 
>statement that __self__ is magical (more or less so doesn't matter 
>much), suggests that together, they are trying to replace an easy to 
>understand magic, with another easy to understand magic.  The only 
>/real/ change that the function unification is advocating, is the 
>replacement of an explicit name for something that is passed implicitly, 
>with an implicit name for something that is passed implicitly, and the 
>replacement of an explicit instance name with no name at all.  Which is 
>better is debatable, but I prefer the current Python syntax, due to the 
>reduced number of implicit items (only the instance is implicit, as 
>compared with the instance, name, and instance references in prototype 
>syntax).

Again, I'm not getting into the 'self' debate.  The real issue is
unification.  If that were to require putting 'self' into *all*
calling sequences, whether it was needed or not, that would be
preferrable to the current mess.

-- Dave