epiphany

[Roy Smith]

I discovered something really neat today.

We've got a system with a bunch of rules. Each rule is a method which
returns True or False. At some point, we need to know if all the rules
are True. Complicating things, not all the rules are implemented.
Those that are not implemented raise NotImplementedError.

We used to have some ugly logic which kept track of which rules were
active and only evaluated those.

So, here's the neat thing. It turns out that bool(NotImplemented)
returns True. By changing the unimplemented rules from raising
NotImplementedError to returning NotImplemented, the whole thing becomes:

return all(r() for r in rules)

 

[Steven D'Aprano]

I discovered something really neat today.

We've got a system with a bunch of rules. Each rule is a method which
returns True or False. At some point, we need to know if all the rules
are True. Complicating things, not all the rules are implemented. Those
that are not implemented raise NotImplementedError.

NotImplementedError is intended to be raised by abstract base classes to
indicate a method that must be overridden. I also use it as a place-
holder for functions or methods I haven't actually written yet. I'm not
sure what semantics you're giving NotImplementedError in your code, but I
wonder whether a neater solution might be to just use rule = None for
unimplemented rules, rather than:

def unimplemented():
raise NotImplementedError

rule = unimplemented

Then your logic for seeing if all rules return true would become:

all(r() for r in rules if r is not None)

and for seeing if all rules return true or are unimplemented:

all(r is None or r() for r in rules)

We used to have some ugly logic which kept track of which rules were
active and only evaluated those.

I don't see why you would need anything like that. Reading further on, I
see that you are counting unimplemented rules as true, for some reason
which I don't understand. (Knowing nothing of your use-case, I would have
expected intuitively that unimplemented rules count as not true.) A
simple helper function will do the job:


def eval(rule):
try:
return rule()
except NotImplementedError:
return True

everything_is_true = all(eval(r) for r in rules)



No need for complicated ugly logic keeping track of what rules are
implemented. But if you're worried about the cost of catching those
exceptions (you've profiled your code, right?) then that's easy with a
decorator:


def not_implemented(func):
@functools.wraps(func)
def inner(*args, **kw):
raise NotImplementedError
inner.ni = True
return inner


# Decorate only the rules you want to be unimplemented.

@not_implemented
def my_rule():
pass


everything_is_true = all(r() for r in rules if not hasattr(r, 'ni'))



Note that if you could reverse the logic so that unimplemented rules
count as not true, this will also work:

try:
everything_is_true = all(r() for r in rules)
except NotImplementedError:
everything_is_true = False

So, here's the neat thing. It turns out that bool(NotImplemented)
returns True. By changing the unimplemented rules from raising
NotImplementedError to returning NotImplemented, the whole thing
becomes:

return all(r() for r in rules)

Objects are supposed to return NotImplemented from special dunder methods
like __add__, __lt__, etc. to say "I don't know how to implement this
method for the given argument". Python will then try calling the other
object's special method. If both objects return NotImplemented, Python
falls back on whatever default behaviour is appropriate.

So, knowing nothing of your application, I fear that this is an abuse of
NotImplemented's semantics. If a rule returns NotImplemented, I would
expect your application to fall back on a different rule. If that's not
the case, you're using it in a non-standard way that will cause confusion
for those with expectations of what NotImplemented means.

 

[Roy Smith]

I don't see why you would need anything like that. Reading further on, I
see that you are counting unimplemented rules as true, for some reason
which I don't understand.

The top-level logic we need to enforce is "this configuration doesn't
violate any rules".

 

[Chris Angelico]

The top-level logic we need to enforce is "this configuration doesn't
violate any rules".

Then have your unimplemented rules simply return True. Easy!

ChrisA

 

[Ethan Furman]

Objects are supposed to return NotImplemented from special dunder methods
like __add__, __lt__, etc. to say "I don't know how to implement this
method for the given argument". Python will then try calling the other
object's special method. If both objects return NotImplemented, Python
falls back on whatever default behaviour is appropriate.

So, knowing nothing of your application, I fear that this is an abuse of
NotImplemented's semantics. If a rule returns NotImplemented, I would
expect your application to fall back on a different rule. If that's not
the case, you're using it in a non-standard way that will cause confusion
for those with expectations of what NotImplemented means.

Why would you assume some random application is going to deal with NotImplemented the same way the python interpreter
does? And even the interpreter isn't consistent -- sometimes it will return false (__eq__) and sometimes it will raise
an Exception (__add__).

I hardly think it an abuse of NotImplemented to signal something is not implemented when NotImplemented means, um, not
implemented.

possibly-not-implemented-ly yours,

 

[Ethan Furman]

Then have your unimplemented rules simply return True. Easy!

And less clear.

 

[Roy Smith]

Then have your unimplemented rules simply return True. Easy!

ChrisA

It's nice to have tri-state logic:

* This rule passes

* This rule fails

* This rule was not evaluated

What I've got now expresses that perfectly.

 

[Steven D'Aprano]

Why would you assume some random application is going to deal with
NotImplemented the same way the python interpreter does?

Why would you assume that some random application is going to treat x==y
the same way the Python interpreter does?

Just because you can design your objects to do anything you want doesn't
mean you should. Breaking conventions carries costs by the mere fact that
you're breaking conventions. There are established semantics that an
experienced Python developer will expect for NotImplemented, and doing
something else risks causing confusion and mistakes.

Or worse, bugs. If there is any chance that a rule might be called in a
context where the Python interpreter gets to interpret the return result
before you see it, then returning NotImplemented could lead to difficult
to debug problems.

And even the
interpreter isn't consistent -- sometimes it will return false (__eq__)
and sometimes it will raise an Exception (__add__).

As I said:

"If both objects return NotImplemented, Python falls back on whatever
default behaviour is appropriate."

If neither object knows how to compare the other for equality, the
appropriate behaviour is to treat them as unequal. If neither object
knows how to add itself to the other, the appropriate behaviour is to
raise an exception.

I hardly think it an abuse of NotImplemented to signal something is not
implemented when NotImplemented means, um, not implemented.

It doesn't just mean "not implemented in general", it has a specific
meaning: "I don't know what to do here, let the other object handle it".

As I have repeatedly said, I don't know the context of the application,
but from what little has been described, this part of it doesn't feel to
me like a good, clean design. I might be wrong, but from the outside it
feels like the API should be that rules return a three-state logic
instance:

True, False, Unknown

where Unknown can be trivially created with

Unknown = object()

The semantics of NotImplementedError is that it is an *error*, and that
doesn't sound appropriate given the example shown. Why would a rule that
raises an *error* exception be treated as if it had passed? That's just
wrong.

The semantics of NotImplemented is that it is a signal for one object to
say "I don't know how to do this, let somebody else try". That also
doesn't seem appropriate. There's no sign that Roy's application does the
equivalent to this:

result = rule()
if result is NotImplemented:
result = another_rule()
if result is NotImplemented:
result = some_default


Since rules apparently take no arguments, either:

1) they rely on global state, which is a nasty design; or

2) rules actually have a fixed return result, in which case why make them
functions in the first place?


Since both possibilities seem stupid, and I do not believe that Roy
actually is stupid, I suspect that his example over-simplifies the
situation. But I can't comment on the infinite number of things that his
code might do, I can only comment on the examples as actually given, and
as given, I don't think that either NotImplementedError or NotImplemented
is a clean solution to the problem.

 

[Roy Smith]

The semantics of NotImplemented is that it is a signal for one object to
say "I don't know how to do this, let somebody else try".

That's precisely the logic here. The rule says, "I don't know how to
tell you if this is OK or not, ask another rule".

Since rules apparently take no arguments, either:

1) they rely on global state, which is a nasty design; or

2) rules actually have a fixed return result, in which case why make them
functions in the first place?

Yes, rules take arguments. I elided them from the original description
since it wasn't germane to what I was trying to show.

Since both possibilities seem stupid, and I do not believe that Roy
actually is stupid,

I am honored that you have such a high opinion of me

Here's what the docs say about NotImplemented:

This type has a single value. There is a single object with this
value. This object is accessed through the built-in name
NotImplemented. Numeric methods and rich comparison methods may
return this value if they do not implement the operation for the
operands provided. (The interpreter will then try the reflected
operation, or some other fallback, depending on the operator.) Its
truth value is true.

It gives an example of a use by numeric methods. It doesn't say that's
the only thing it can be used for.

It also says, "Its truth value is true". Why would they document that
fact if you weren't supposed to use it as a boolean operand?

 

[Steven D'Aprano]

That's precisely the logic here. The rule says, "I don't know how to
tell you if this is OK or not, ask another rule".

Sounds good to me then. It looks like your design is actually much closer
to what I believe the standard Python semantics are intended to be than
it appeared at first.



[...]

Here's what the docs say about NotImplemented:


It gives an example of a use by numeric methods. It doesn't say that's
the only thing it can be used for.

Right. You can do a lot of things in Python, including shooting your foot
off but that doesn't mean you should. The further away from standard
Python conventions you get, the more wary you should be. That's all.

It also says, "Its truth value is true". Why would they document that
fact if you weren't supposed to use it as a boolean operand?

You can use *anything* in Python in a boolean context. That's a language
feature: all objects are either truthy or falsey. As for why it is
documented for NotImplemented, I guess that's because some people might
guess that it is falsey, like None.

 

[Roy Smith]

It also says, "Its truth value is true". Why would they document that
fact if you weren't supposed to use it as a boolean operand?

You can use *anything* in Python in a boolean context. That's a language
feature: all objects are either truthy or falsey. As for why it is
documented for NotImplemented, I guess that's because some people might
guess that it is falsey, like None.[/QUOTE]

That was part of what added the epiphanality to the experience. My
first guess was exactly as you say, that bool(NotImplemented) would be
false. Once I discovered that it was true, the rest immediately fell
into place and many lines of code got replaced by the simple:

return all(r(...) for r in rules)
^
|
+---- stuff that I'm not showing goes here

 

[88888 Dihedral]

Roy Smithæ–¼ 2013å¹´4月25日星期四UTC+8上åˆ7時50分33秒寫é“：

I discovered something really neat today.



We've got a system with a bunch of rules. Each rule is a method which

returns True or False. At some point, we need to know if all the rules

are True. Complicating things, not all the rules are implemented.

Those that are not implemented raise NotImplementedError.



We used to have some ugly logic which kept track of which rules were

active and only evaluated those.



So, here's the neat thing. It turns out that bool(NotImplemented)

returns True. By changing the unimplemented rules from raising

NotImplementedError to returning NotImplemented, the whole thing becomes:



return all(r() for r in rules)

Problems of rules in Boolean algebra or the bi-level logic
inference engine in AI were all solved long time ago
in the text book about AI.

There are some variations about the multi-level or
the continuous level logic engine with some new phases
in Fuzzy theory in the expert system.

A dynamical typed language is better to be used in this kind of
problems.