Sentinel values for special cases

[Ben Finney]

Howdy all,

Ned Batchelder blogged[0] about a debate over checking function
parameters, and to what extent dynamic typing should be relied upon.

I was one of many who commented, but I wrote what purports to be a
comprehensive checklist when considering special-case inputs for
functions. I thought the best way to puncture my ego would be to post
that list here and see how well it survives :-) Here goes:


If you have a special case, there are a few steps.

First, do you really need a special case, or are you just being
paranoid about type safety? Let the caller take care of whether they
mean what they say, and make your function do *one* job clearly and
simply. (This supports the 'insert_ids(get _global_ids())' idea
earlier.)

Second, do you *really* need a special case, or are you making your
function too complex? Be very suspicious of functions that are written
to do two different things depending on their input, and split them so
that both are simple and the caller can be explicit about what they
want. This doesn't preclude factoring out the code that's common to
both of them, of course.

Third, if you actually need a special case, can it be None? This is
the idiomatic Python "sentinel value", and it looks like the code
posted by 'sri' [reproduced below].

def insert_ids(ids= None):
if ids is None:
ids = get_global_ids( )

Note that if you're squeamish about using None,
but don't have a specific reason not to use it, use it; other
programmers will thank you for following convention.

Fourth, if you have decided that a magic sentinel value is called for
but None is already taken for some other purpose, don't use a
string. Use a unique do-nothing object, defined at the module level so
callers can easily get at it, like 'Dmitry Vasiliev' showed
[reproduced below].

GLOBAL = object()

def insert_ids(ids= GLOBAL):
if ids is GLOBAL:
ids = get_global_ids( )

You won't accidentally use it, because it's defined only in one place
(you're comparing by 'is', remember) and it's not used for anything
except indicating the special case.

Fifth, there is no fifth. If you've come to the end and think it's too
complex, it probably is. Start at the top again.



[0]: <URL:http://www.nedbatcheld er.com/blog/200610.html#e20 061022T192641>

--
\ "Immorality : The morality of those who are having a better |
`\ time." -- Henry L. Mencken |
_o__) |
Ben Finney

[Paddy]

Re: Sentinel values for special cases

Ben Finney wrote:

Howdy all,
>
Ned Batchelder blogged[0] about a debate over checking function
parameters, and to what extent dynamic typing should be relied upon.
>
I was one of many who commented, but I wrote what purports to be a
comprehensive checklist when considering special-case inputs for
functions. I thought the best way to puncture my ego would be to post
that list here and see how well it survives :-) Here goes:
>
[0]: <URL:http://www.nedbatcheld er.com/blog/200610.html#e20 061022T192641>

Referring to the blog entry, if I had a function that could take
several items of data, where data items were 'expensive' to gather, but
might not all be needed due to other data values, then I can see the
need for using sentinel values for some of the data items and only
computing their true value when necessary.

In Python, as others stated, the sentinel value of choice for all the
data values should be None, or, some module level global constant,
guaranteed not to be part of the data.

None is what Python readers would expect as a sentinel value, but if
any of your data fields could have None as a valid value then you may
have to switch to a module level constant.
Be wary of using sentinel values which are strings, if your data could
itself be a string - make sure the sentinel value is not valid data,
and always use the sentinel name and not its value from then on. it is
very wrong to do this sort of thing:


NO_DATA = '::NO_DATA::'

def xyz(a,b,c):
if a == '::NO_DATA::':
# blah blah blah


You should use the name NO_DATA for the comparison.

If you are having difficulty working out what to use as a sentinel
value for your data then you could declare a Sentinel class and use
that:


class _Sentinel(objec t):
' Initial data value when true data has not been fetched/computed
yet'
pass

NO_DATA = _Sentinel

def xyz(a,b,c):
if a == NO_DATA:
# go get a


(Hmm, should that be double underscores on _Sentinel ...).

- Paddy.

[Ben Finney]

Re: Sentinel values for special cases

"Paddy" <paddy3118@nets cape.netwrites:

None is what Python readers would expect as a sentinel value, but if
any of your data fields could have None as a valid value then you may
have to switch to a module level constant.
Be wary of using sentinel values which are strings, if your data could
itself be a string - make sure the sentinel value is not valid data,
and always use the sentinel name and not its value from then on.

I don't think "be wary" is appropriate; I think a blanket "don't use
strings for sentinel values" is fine. In your example below:

it is very wrong to do this sort of thing:
>
NO_DATA = '::NO_DATA::'
>
def xyz(a,b,c):
if a == '::NO_DATA::':
# blah blah blah

You should not use a string for that constant at all. Its only purpose
is to be a sentinel value, so just make it a unique object instance:

NO_DATA = object()

def xyz(a,b,c):
if a is NO_DATA:
# blah blah blah

The sentinel value for the 'a' parameter to 'foo.xyz()' is then
'foo.NO_DATA', nothing else. We're now checking by 'is', not '==', so
there's no way to pass the sentinel value to the function unless it's
that specific object. It's impossible to get that particular value any
other way, so you avoid the possibility of accidentally getting the
sentinel as part of the data.

I don't see any advantage a string would have to make it worth using
over the above type of sentinel.

If you are having difficulty working out what to use as a sentinel
value for your data then you could declare a Sentinel class and use
that:

Just use the plain 'object' class, since the instance name will make
its meaning clear, and other instances don't need to share any
behaviour or semantics.

--
\ "There is no reason anyone would want a computer in their |
`\ home." -- Ken Olson, president, chairman and founder of |
_o__) Digital Equipment Corp., 1977 |
Ben Finney

[Aahz]

Re: Sentinel values for special cases

In article <mailman.1308.1 161912647.11739 .python-list@python.org >,
Ben Finney <ben@benfinney. id.auwrote:

>
>Fourth, if you have decided that a magic sentinel value is called for
>but None is already taken for some other purpose, don't use a
>string. Use a unique do-nothing object, defined at the module level so
>callers can easily get at it, like 'Dmitry Vasiliev' showed
>[reproduced below].
>
GLOBAL = object()
>
def insert_ids(ids= GLOBAL):
if ids is GLOBAL:
ids = get_global_ids( )

The one disadvantage of this approach is that it complicates pickling
if/when you store the stentinel in an instance. There are ways of
working around that, but none are pleasant.
--
Aahz (aahz@pythoncra ft.com) <* http://www.pythoncraft.com/

"If you don't know what your program is supposed to do, you'd better not
start writing it." --Dijkstra

[Gabriel Genellina]

Re: Sentinel values for special cases

At Friday 27/10/2006 11:40, Aahz wrote:

GLOBAL = object()

def insert_ids(ids= GLOBAL):
if ids is GLOBAL:
ids = get_global_ids( )

>
>The one disadvantage of this approach is that it complicates pickling
>if/when you store the stentinel in an instance. There are ways of
>working around that, but none are pleasant.

But why should you store the sentinel in an instance? It's only
purpose is to detect a special case in the parameter, when None is
not appropiate.
Even if you were assigning instance attributes, you can use the
sentinel as a default class attribute (which are not pickled).

>>GLOBAL = object()
>>class A(object):

.... x = GLOBAL
.... def __init__(self, x=GLOBAL):
.... if x is not GLOBAL:
.... self.x = x
....

>>a = A(1)
>>a.x

1

>>vars(a)

{'x': 1}

>>b = A()
>>b.x

<object object at 0x00B8B438>

>>vars(b)

{}

>>>
>>from cPickle import loads,dumps
>>s=dumps(b)
>>b2=loads(s)
>>b2.x

<object object at 0x00B8B438>

>>vars(b2)

{}



--
Gabriel Genellina
Softlab SRL

_______________ _______________ _______________ _____
Correo Yahoo!
Espacio para todos tus mensajes, antivirus y antispam ¡gratis!
¡Abrí tu cuenta ya! - http://correo.yahoo.com.ar

[Ben Finney]

Re: Sentinel values for special cases

aahz@pythoncraf t.com (Aahz) writes:

Ben Finney <ben@benfinney. id.auwrote:

Use a unique do-nothing object, defined at the module level so
callers can easily get at it [...]

GLOBAL = object()

def insert_ids(ids= GLOBAL):
if ids is GLOBAL:
ids = get_global_ids( )

>
The one disadvantage of this approach is that it complicates
pickling if/when you store the stentinel in an instance. There are
ways of working around that, but none are pleasant.

Hmm, and any other kind of serialisation would be similarly affected I
suppose.

What's an alternative?

--
\ "Think for yourselves and let others enjoy the privilege to do |
`\ so too." -- Voltaire, _Essay On Tolerance_ |
_o__) |
Ben Finney

[Ben Finney]

Re: Sentinel values for special cases

Gabriel Genellina <gagsl-py@yahoo.com.ar writes:

At Friday 27/10/2006 11:40, Aahz wrote:

Ben Finney wrote:

GLOBAL = object()
>
def insert_ids(ids= GLOBAL):
if ids is GLOBAL:
ids = get_global_ids( )

The one disadvantage of this approach is that it complicates
pickling if/when you store the stentinel in an instance. There are
ways of working around that, but none are pleasant.

>
But why should you store the sentinel in an instance? It's only
purpose is to detect a special case in the parameter, when None is
not appropiate.

You might be storing values that will later become parameters to the
function.

ids_to_be_proce ssed_later = {
'chin': [14, 7, 9],
'bin': foo.GLOBAL,
'fin': [74, 98, 12],
}
serialise_stuff ()

--
\ "The only tyrant I accept in this world is the still voice |
`\ within." -- Mahatma Gandhi |
_o__) |
Ben Finney

[Gabriel Genellina]

Re: Sentinel values for special cases

At Friday 27/10/2006 21:09, Ben Finney wrote:

GLOBAL = object()

def insert_ids(ids= GLOBAL):
if ids is GLOBAL:
ids = get_global_ids( )
>
>The one disadvantage of this approach is that it complicates
>pickling if/when you store the stentinel in an instance. There are
>ways of working around that, but none are pleasant.

But why should you store the sentinel in an instance? It's only
purpose is to detect a special case in the parameter, when None is
not appropiate.

>
>You might be storing values that will later become parameters to the
>function.
>
ids_to_be_proce ssed_later = {
'chin': [14, 7, 9],
'bin': foo.GLOBAL,
'fin': [74, 98, 12],
}
serialise_stuff ()

Well... just don't do that! :)
Instead of a dict, use another class instance with a class attribute
(like on my earlier post).


--
Gabriel Genellina
Softlab SRL

_______________ _______________ _______________ _____
Correo Yahoo!
Espacio para todos tus mensajes, antivirus y antispam ¡gratis!
¡Abrí tu cuenta ya! - http://correo.yahoo.com.ar