Why are there no ordered dictionaries?
Christoph Zwerschke
haven't found a satisfying answer yet: Why isn't there an "ordered
dictionary" class at least in the standard list? Time and again I am
missing that feature. Maybe there is something wrong with my programming
style, but I rather think it is generally useful.
I fully agree with the following posting where somebody complains why so
very basic and useful things are not part of the standard library:
http://groups.google.com/group/comp.lang.python/msg/e652d2f771a49857
Are there plans to get it into the standard lib sometime?
-- Christoph
Ben Finney
> This is probably a FAQ, but I dare to ask it nevertheless since I
> haven't found a satisfying answer yet: Why isn't there an "ordered
> dictionary" class at least in the standard list?
What answers have you received that have not been satisfactory?
Some possible answers: The native dict is very fast, partly because
the implementation doesn't need to ensure any particular ordering.
Ordering the keys isn't the normal case, and can be done easily when
needed.
You asked "why not" rather than "has anyone done this anyway"; if
you asked the latter of the Python Cookbook, you might find something
like this:
<URL:http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/107747>
A little old, and pre-dates subclassable native types, but quite
serviceable.
> Time and again I am missing that feature. Maybe there is something
> wrong with my programming style, but I rather think it is generally
> useful.
In what cases do you find yourself needing a dict that preserves its
key order? Can you present a use case that would be improved by an
ordered dict?
> I fully agree with the following posting where somebody complains
> why so very basic and useful things are not part of the standard
> library:
For my part, I consider it a virtue of Python that the standard
library doesn't change rapidly. It allows many competing
implementations to be shaken out before everyone starts depending on
any one of them.
> Are there plans to get it into the standard lib sometime?
Where to find an answer:
<URL:http://www.python.org/peps/pep-0000.html>
Where to change that answer:
<URL:http://www.python.org/peps/pep-0001.html>
--
\ "One of the most important things you learn from the internet |
`\ is that there is no 'them' out there. It's just an awful lot of |
_o__) 'us'." -- Douglas Adams |
Ben Finney
przemek drochomirecki
Uzytkownik "Christoph Zwerschke" <ci...@online.de> napisal w wiadomosci
news:dlpr88$ts6$1...@online.de...
i am not sure what is the purpose of having ordered dictionaries built in
python, could u provide any examples?
i use a contruction:
for x in sorted(d.keys())
cheers,
przemek
bon...@gmail.com
przemek drochomirecki wrote:
> i am not sure what is the purpose of having ordered dictionaries built in
> python, could u provide any examples?
>
> i use a contruction:
> for x in sorted(d.keys())
>
be derived from the content, say insertion order(most ordered dict I
saw use this order).
I am writing a web applications(simple forms) which has a number of
fields. Each field naturally has a name and a number of
attributes(formatting etc.), like this :
d = {'a':{...}, 'b':{....}}
This dict would be passed to the Kid template system which would lay it
out into a HTML form. For quick and dirty forms, I don't want to code
each field individually in the HTML template but just from top to
bottom(or left to right for a table) with a for loop.
However, I still want to group certain fields together. This is my need
of an ordered dict. Or course, I can pass a list along with the dict
and loop over the list and retrieve value from the dict, but that would
mean another things to pass along. And given the constraint of Kid
where everything must be one-liner(expression, no block of code), it
makes thing a bit harder.
Fredrik Lundh
> I am writing a web applications(simple forms) which has a number of
> fields. Each field naturally has a name and a number of
> attributes(formatting etc.), like this :
>
> d = {'a':{...}, 'b':{....}}
>
> This dict would be passed to the Kid template system which would lay it
> out into a HTML form. For quick and dirty forms, I don't want to code
> each field individually in the HTML template but just from top to
> bottom(or left to right for a table) with a for loop.
>
> However, I still want to group certain fields together. This is my need
> of an ordered dict.
huh? if you want a list, use a list.
d = [('a', {...}), ('b', {....})]
</F>
bon...@gmail.com
a list ? But the same template, it would(may) be further enhanced by
graphic designers in which case, I need direct access to the field
names, thus the dict property.
In this way, I don't have to change the python code just because I
change the presentation in the template.
Nicola Larosa
> you asked the latter of the Python Cookbook, you might find something
> like this:
>
> <URL:http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/107747>
>
> A little old, and pre-dates subclassable native types, but quite
> serviceable.
Here's a more recent and tested one, by yours truly (and Michael Foord):
An Ordered Dictionary
http://www.voidspace.org.uk/python/odict.html
--
Nicola Larosa - nico...@m-tekNico.net
How wonderful the world would be if his behaviour and attitude was the
default among rich people - using his money with a vision to improve the
world, instead of getting 8 sportcars and a larger penis.
-- barkholt on Slashdot, October 2005, referring to Mark Shuttleworth
Christoph Zwerschke
I googled a little bit and haven't found many answers at all. Some were
in the posting I mentioned: Stuff should go into the standard lib only
when it is mature and "right" enough. However, we are already at Python
2.4 and there is still no ordered dictionary, though there is a lot of
other specialized stuff.
> Some possible answers: The native dict is very fast, partly because
> the implementation doesn't need to ensure any particular ordering.
Ok, but that's not an argument against providing ordered dictionaries as
well.
> Ordering the keys isn't the normal case, and can be done easily when
> needed.
That depends. Maybe I do not want the keys to be sorted alphabetically,
but according to some criteria which cannot be derived from the keys
themselves.
> You asked "why not" rather than "has anyone done this anyway"; if
> you asked the latter of the Python Cookbook, you might find something
> like this.
Yes, I also found that others have done it more than once, and I know
that it's not so difficult to do. There are at least two recipes in the
mentioned cookbook and there is odict in pythonutils. The question was
why is it not available in the *standard* lib.
> In what cases do you find yourself needing a dict that preserves its
> key order? Can you present a use case that would be improved by an
> ordered dict?
There are too many different situations and it would be too much to
explain them here, usually in the case mentioned above where the keys
are not sorted alphabetically. I often solve them by using two data
structures, a list or tuple, plus a dictionary. For instance, the list
could contain the names of database fields which shall be output in a
certain order, and the dictionary values could contain human readable
description of the database fields for headers or something. Instead of
maintaining both data structures I feel it would be more natural to use
only one ordered dictionary.
> For my part, I consider it a virtue of Python that the standard
> library doesn't change rapidly. It allows many competing
> implementations to be shaken out before everyone starts depending on
> any one of them.
Ok, but this can be used as an argument to not add anything to the
standard lib any more. There are already enough competing
implementations. Also, the implementation details are not so important,
there must be only agreement on the interface and behavior which should
not be so difficult in this case.
I simply wanted to ask why it is not available in the standard lib,
since I simply don't know
- has it not been demanded loud enough?
- is it really not needed (if you need it it shows you are doing
something wrong)?
- because nobody presented a satisfying implementation yet?
- are there hidden difficulties or controversial issues?
-- Christoph
Christoph Zwerschke
> By ordered dict, one usually wants order that is arbitary which cannot
> be derived from the content, say insertion order(most ordered dict I
> saw use this order).
> I am writing a web applications(simple forms) which has a number of
> fields. Each field naturally has a name and a number of
> attributes(formatting etc.), like this :
> d = {'a':{...}, 'b':{....}}
Things like that are also my typical use case. The keys usually contain
things like database fields or html form fields, the values contain the
corresponding description, formatting, data type or data itself etc.
The example above is a bit misleading, because using 'a', 'b' as keys
can give the impression that you just have to sort() the keys to have
what you want. So let's make it more realistic:
d = { 'pid': ('Employee ID', 'int'),
'name': ('Employee name', 'varchar'),
'sal': ('Salary', 'float') }
Now if I want these things to be presented in this order, I need to run
through a separate list ('pid', 'name', 'sal') that holds the order.
Ok, you could simply use a list instead of a dictionary:
d = ( ('pid', 'Employee ID', 'int'),
('name', 'Employee name', 'varchar'),
('sal', 'Salary', 'float') )
This works as long as you *only* have to go through the list
sequentially. But maybe you want to print the name with its description
at some other place as well. Now how do you access its description
'Employee name' so easily?
-- Christoph
Ben Finney
Christoph Zwerschke <ci...@online.de> wrote:
> Ben Finney wrote:
> > In what cases do you find yourself needing a dict that preserves
> > its key order? Can you present a use case that would be improved
> > by an ordered dict?
>
> There are too many different situations and it would be too much to
> explain them here, usually in the case mentioned above where the
> keys are not sorted alphabetically.
Without an example, it's hard to know what you want to do and whether
an ordered dictionary is the best way to do it.
> > For my part, I consider it a virtue of Python that the standard
> > library doesn't change rapidly. It allows many competing
> > implementations to be shaken out before everyone starts depending
> > on any one of them.
>
> Ok, but this can be used as an argument to not add anything to the
> standard lib any more.
I hope not. Rather, it's an argument not to add something to the
standard library until it's proven (to the BDFL's criteria) that it's
better in than out.
> There are already enough competing
> implementations.
Have they been sufficiently shaken out to show a clearly superior
version? Is any version sufficiently beneficial to write a PEP for its
inclusion in the standard library?
> I simply wanted to ask why it is not available in the standard lib,
> since I simply don't know
>
> - has it not been demanded loud enough?
Loud demands don't count for much. PEPs with popular working
implementations do.
> - is it really not needed (if you need it it shows you are doing
> something wrong)?
You dismissed a request for your use cases with handwaving. How can we
know?
> - because nobody presented a satisfying implementation yet?
I'm not sure what you mean by "satisfying".
> - are there hidden difficulties or controversial issues?
Another possibility: ordered dictionaries are not needed when Python
2.4 has the 'sorted' builtin.
--
\ "Those who will not reason, are bigots, those who cannot, are |
`\ fools, and those who dare not, are slaves." -- "Lord" George |
_o__) Gordon Noel Byron |
Ben Finney
Fredrik Lundh
> The example above is a bit misleading, because using 'a', 'b' as keys
> can give the impression that you just have to sort() the keys to have
> what you want. So let's make it more realistic:
>
> d = { 'pid': ('Employee ID', 'int'),
> 'name': ('Employee name', 'varchar'),
> 'sal': ('Salary', 'float') }
>
> Now if I want these things to be presented in this order, I need to run
> through a separate list ('pid', 'name', 'sal') that holds the order.
>
> Ok, you could simply use a list instead of a dictionary:
>
> d = ( ('pid', 'Employee ID', 'int'),
> ('name', 'Employee name', 'varchar'),
> ('sal', 'Salary', 'float') )
if you restructure the list somewhat
d = (
('pid', ('Employee ID', 'int')),
('name', ('Employee name', 'varchar')),
('sal', ('Salary', 'float'))
)
you can still loop over the list
for key, (name, type) in d:
print key, name, type # e.g. generate form entry
> This works as long as you *only* have to go through the list
> sequentially. But maybe you want to print the name with its description
> at some other place as well. Now how do you access its description
> 'Employee name' so easily?
but you can easily generate an index when you need it:
index = dict(d)
name, type = index["pid"]
print name
the index should take less than a microsecond to create, and since it
points to the members of the original dict, it doesn't use much memory
either...
</F>
bon...@gmail.com
Fredrik Lundh wrote:
> but you can easily generate an index when you need it:
>
> index = dict(d)
>
> name, type = index["pid"]
> print name
>
> the index should take less than a microsecond to create, and since it
> points to the members of the original dict, it doesn't use much memory
> either...
>
as we can always convert a string field into some other types when it
is needed.
Of course there are more than one way to skin a cat(well it may be
against the general idiom of python) but in some situation certain way
is preferred.
Christoph Zwerschke
> if you restructure the list somewhat
> d = (
> ('pid', ('Employee ID', 'int')),
> ('name', ('Employee name', 'varchar')),
> ('sal', ('Salary', 'float'))
> )
> you can still loop over the list
> but you can easily generate an index when you need it:
> index = dict(d)
That's exactly the kind of things I find myself doing too often and what
I was talking about: You are using *two* pretty redundant data
structures, a dictionary and a list/tuple to describe the same thing.
Ok, you can use a trick to automatically create the dictionary from the
tuple, but still it feels somewhat "unnatural" for me. A "ordered
dictionary" would be the more "natural" data structure here.
I also wanted to mention the uglyness in the definition (nested tuples),
but then I understood that even an ordered dictionary would not
eliminate that uglyness, since the curly braces are part of the Python
syntax and cannot be used for creating ordered dictionaries anyway. I
would have to define the ordered dictionary in the very same ugly way:
d = odict(('pid', ('Employee ID', 'int')),
('name', ('Employee name', 'varchar')),
('sal', ('Salary', 'float')))
(Unless the Python syntax would be extend to use double curly braces or
something for ordered dictionaries - but I understand that this is not
an option.)
-- Christoph
bon...@gmail.com
Ben Finney wrote:
> Another possibility: ordered dictionaries are not needed when Python
> 2.4 has the 'sorted' builtin.
>
or some arbitary order that instead of a,b,c,d,e, I want it as e, c, b,
d, a ?
Personally, I have needs for ordered dict but I don't think it should
be in standard library though, as different situation called for
different behaviour for "ordered" and skewing my code to a standard lib
way is no good.
What I think is better is like the itertools recipe of giving example
of how one can make their own based on the needs.
Christoph Zwerschke
> Without an example, it's hard to know what you want to do and whether
> an ordered dictionary is the best way to do it.
I have indicated an example, discussed in more detail in another subthread.
>> There are already enough competing implementations.
> Have they been sufficiently shaken out to show a clearly superior
> version? Is any version sufficiently beneficial to write a PEP for its
> inclusion in the standard library?
At least it shows I'm not the only one who thinks ordered dictionaries
may be sometimes nice to have.
>> I simply wanted to ask why it is not available in the standard lib,
>> since I simply don't know
>> - has it not been demanded loud enough?
> Loud demands don't count for much. PEPs with popular working
> implementations do.
Sorry, I did not mean "loud enough" but "often enough". The same what
you are calling "popular."
>> - because nobody presented a satisfying implementation yet?
> I'm not sure what you mean by "satisfying".
You can take your own definition: "sufficiently shaken out", "working",
"popular", and "succifiently beneficial" and "proven (to the BDFL's
criteria)".
> Another possibility: ordered dictionaries are not needed when Python
> 2.4 has the 'sorted' builtin.
The 'sorted' function does not help in the case I have indicated, where
"I do not want the keys to be sorted alphabetically, but according to
some criteria which cannot be derived from the keys themselves."
-- Christoph
Christoph Zwerschke
> Personally, I have needs for ordered dict but I don't think it should
> be in standard library though, as different situation called for
> different behaviour for "ordered" and skewing my code to a standard lib
> way is no good.
I have started the thread in the first place because I believed it is
pretty unabmiguous what an "ordered dictionary" is and how it should
behave. That's why I asked myself why something that straigthforward has
not been added to the standard lib yet. Maybe I'm wrong; I must admit
that I haven't meditated about it very much.
Do you have an example for different options of behavior?
-- Christoph
bon...@gmail.com
assume that is the need of their creators. But that is not my need, so
there are at least two behaviour. What I need is a "preferred order".
Say if I have designed a web form(correspond to a database table), I
just want say 3 fields that goes before anything else in the
presentation. The rest I don't care as the DBA may create more fields
later which I don't want to then update my code yet again.
Alex Martelli
...
> I have started the thread in the first place because I believed it is
> pretty unabmiguous what an "ordered dictionary" is and how it should
I think you're wrong here. People in the past who have requested or
implemented stuff they called 'ordered dicts' in the past had in mind
drastically different things, based on some combination of insertion
orders, keys, and _values_. So, ambiguity is definitely present in the
phrase 'ordered dictionary', because there are so many different
criteria whereby the 'ordering' could take place.
Note the plural in 'insertion orderS': some people care about the FIRST
time a key was added to a dict, some about the LAST time it was added,
some about the latest time it was 'first inserted' (added and wasn't
already there) as long as it's never been deleted since that occasion --
and these are just a few of the multifarious orders based on the time of
insertions and deletions of keys. The number of variations is
staggering, e.g., consider
x['a'] = 1
x['b'] = 2
x['a'] = 1
in some applications you'd want to have 'b' come before 'a' because the
last time of addition was earlier for 'b' -- but in others you might
want 'a' first because the latest addition wasn't really one, since it
didn't really change anything (because the value inserted was the same
as the one already there -- it would be different, for those other apps,
if the RHS of the third assignment was 0 rather than 1...).
To get 'ordered dicts' into Python, you have to identify ONE unambiguous
definition which has a large-enough number of use-cases, possibly
customizable through some reasonably SIMPLE combination of flags and a
callable or two, like the 'sorted' built-in has a 'reversed' flag and
'key' and 'cmp' optional callables. Expect a lot of flak from those who
have been pining for an 'ordered dict' which does NOT match your one
unambiguous definition...;-)
If the field of use cases for 'ordered dicts' is just too fragmented,
it's quite possible that it's best not to have any single kind built-in,
even though, could all different use cases be combined (which by
hypothesis is unfeasible), "critical mass" would be reached...
Alex
Alex Martelli
...
> there are at least two behaviour. What I need is a "preferred order".
> Say if I have designed a web form(correspond to a database table), I
> just want say 3 fields that goes before anything else in the
> presentation. The rest I don't care as the DBA may create more fields
> later which I don't want to then update my code yet again.
preferred_fields = ['foo', 'bar', 'baz']
def process_preferred_fields():
global preferred_fields
temp = {}
for i, field in enumerate(preferred_fields):
temp[field] = '%s%s' % (chr(0), chr(i))
preferred_fields = temp
process_preferred_fields()
del process_preferred_fields
def sort_key(akey, preferred_fields=preferred_fields):
return preferred_fields.get(akey, akey)
del preferred_fields
## ...build dictionary d...
# now output d...:
for k in sorted(d, key=sort_key):
print k, d[k]
Season to taste if you want non-preferred fields emitted other than
alphabetically, or if you want to wrap this stuff into a class, etc.
(Note: untested code, so typos &c are quite possible). This assumes
that no 'real' key is a non-string, and no 'real' key starts with
chr(0), but it's quite easy to tweak for slightly different specs (at
worst by defining a custom type designed to always compare less than any
real key, and wrapping the preferred_fields entry in instances of that
custom type... having such instances compare with each other based on
the index within preferred_fields of the key they're wrapping, etc etc).
Alex
Alex Martelli
> "I do not want the keys to be sorted alphabetically, but according to
> some criteria which cannot be derived from the keys themselves."
Ah, but WHAT 'some criteria'? There's the rub! First insertion, last
insertion, last insertion that wasn't subsequently deleted, last
insertion that didn't change the corresponding value, or...???
Alex
Mike Meyer
> Ben Finney wrote:
>> Another possibility: ordered dictionaries are not needed when Python
>> 2.4 has the 'sorted' builtin.
> The 'sorted' function does not help in the case I have indicated,
> where "I do not want the keys to be sorted alphabetically, but
> according to some criteria which cannot be derived from the keys
> themselves."
And how would an ordered dictionary help in this case?
<mike
--
Mike Meyer <m...@mired.org> http://www.mired.org/home/mwm/
Independent WWW/Perforce/FreeBSD/Unix consultant, email for more information.
bon...@gmail.com
d = somedict_from_db()
prefer=['f','a',b']
def my_order(d):
for x in prefer:
if x in d: yield x
s = frozenset(prefer)
for x in d:
if x not in s: yield x
Peter Hansen
> Using the same logic, we don't need types other than string in a DBMS
> as we can always convert a string field into some other types when it
> is needed.
You mean, like SQLite does? (http://www.sqlite.org/datatypes.html)
-Peter
bon...@gmail.com
Bengt Richter
>> Ordering the keys isn't the normal case, and can be done easily when
>> needed.
>
>That depends. Maybe I do not want the keys to be sorted alphabetically,
>but according to some criteria which cannot be derived from the keys
>themselves.
sorted(plaindict.items(), key=your_ordering_function) ?
>>> def show(*a): print a
...
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=show)
(('a', 97),)
(('c', 99),)
(('b', 98),)
(('d', 100),)
[('a', 97), ('c', 99), ('b', 98), ('d', 100)]
What key function would you like, to generate the value that is actually used
to define the ordering?
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:t[0])
[('a', 97), ('b', 98), ('c', 99), ('d', 100)]
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:t[1])
[('a', 97), ('b', 98), ('c', 99), ('d', 100)]
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:-t[1])
[('d', 100), ('c', 99), ('b', 98), ('a', 97)]
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:t[1]&1)
[('b', 98), ('d', 100), ('a', 97), ('c', 99)]
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:(t[1]&1,t[1]))
[('b', 98), ('d', 100), ('a', 97), ('c', 99)]
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:(t[1]&1,-t[1]))
[('d', 100), ('b', 98), ('c', 99), ('a', 97)]
And being able to reverse the end result is handy
>>> sorted(dict((c,ord(c)) for c in 'abcd').items(), key=lambda t:(t[1]&1,-t[1]), reverse=True)
[('a', 97), ('c', 99), ('b', 98), ('d', 100)]
You may need to upgrade your Python though ;-)
Regards,
Bengt Richter
bon...@gmail.com
Bengt Richter wrote:
> On Sun, 20 Nov 2005 22:03:34 +0100, Christoph Zwerschke <ci...@online.de> wrote:
> >> Ordering the keys isn't the normal case, and can be done easily when
> >> needed.
> >
> >That depends. Maybe I do not want the keys to be sorted alphabetically,
> >but according to some criteria which cannot be derived from the keys
> >themselves.
> You mean involving also the values? What's wrong with
> sorted(plaindict.items(), key=your_ordering_function) ?
>
other words, some other metadata that is not present in the data. A
typical thing, like order of creation. Or some arbitary order. For
example :
I present a data grid/table in a HTML form and the user just drag and
drop and rearrange the columns order.
Of course, you may say, just put another column that represent
this(some reporting programs I have seen do it this way) and that is an
option but not the only option.
Ben Finney
> [sort by] some other metadata that is not present in the data.
> [...]
> Of course, you may say, just put another column that represent
> this(some reporting programs I have seen do it this way) and that is
> an option but not the only option.
It's a pretty good option, and very commonly used. It's known as the
"Schwartzian transform", or more descriptively, the "Decorate, Sort,
Undecorate" pattern.
<URL:http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/52234>
--
\ "Experience is that marvelous thing that enables you to |
`\ recognize a mistake when you make it again." -- Franklin P. |
_o__) Jones |
Ben Finney
bon...@gmail.com
Ben Finney wrote:
> bon...@gmail.com <bon...@gmail.com> wrote:
> > [sort by] some other metadata that is not present in the data.
> > [...]
> > Of course, you may say, just put another column that represent
> > this(some reporting programs I have seen do it this way) and that is
> > an option but not the only option.
>
> It's a pretty good option, and very commonly used. It's known as the
> "Schwartzian transform", or more descriptively, the "Decorate, Sort,
> Undecorate" pattern.
>
is the one seeing the whole thing, and not some snippet(or concept) on
the usenet.
Fredrik Lundh
> Fredrik Lundh wrote:
> > but you can easily generate an index when you need it:
> >
> > index = dict(d)
> >
> > name, type = index["pid"]
> > print name
> >
> > the index should take less than a microsecond to create, and since it
> > points to the members of the original dict, it doesn't use much memory
> > either...
> >
> Using the same logic, we don't need types other than string in a DBMS
> as we can always convert a string field into some other types when it
> is needed.
No, that's not the same logic. The dict() in my example doesn't convert be-
tween data types; it provides a new way to view an existing data structure.
There's no parsing involved, nor any type guessing. And given the use case,
it's more than fast enough, and doesn't copy any data.
If you think that's the same thing as parsing strings, you've completely missed
the point.
</F>
bon...@gmail.com
Well, forget about the missing/not missing the point.
My point is, there are various of reasons why we need different data
types in an RDBMS, just the same as why we need list, dict. There is
nothing stop me from using a list as dict(just scan it till I find it),
why would I I create a dict(your new view of the same data) ? Coding
convenience, speed or whatever.
If I need the dict feature 90% of the time, and the list feature 10% of
the time. I want an ordered dict. Rather than a list and create this
new view every time and every where I want to use it as a dict.
parsing or not parsing is not the point, and parsing/converting is
still "create a new view" of an existing data structure.
Fredrik Lundh
> If I need the dict feature 90% of the time, and the list feature 10% of
> the time.
Wasn't your use case that you wanted to specify form fields in
a given order (LIST), render a default view of the form in that
order (LIST), and, later on, access the field specifiers in an
arbitrary order, based on their key (DICT). Sure looks like it's
the LIST aspect that's important here...
("but assume that I have some other use case" isn't a valid use
case)
> I want an ordered dict. Rather than a list and create this new view every
> time and every where I want to use it as a dict.
You want an ordered dict because you say you want one, not be-
cause it's the best way to address your use case. That's fine, but
it's not really related to the question asked in the subject line.
> parsing or not parsing is not the point, and parsing/converting is
> still "create a new view" of an existing data structure.
Copying the entire data structure hardly qualifies as "creating a
new view". dict() doesn't do that; in this use case, it doesn't cost
you anything to use it.
Everything has a cost in Python. Things aren't free just because
they're implemented by some other module.
But when things are free, they're often a good choice.
</F>
bon...@gmail.com
Fredrik Lundh wrote:
> bon...@gmail.com wrote:
>
> > If I need the dict feature 90% of the time, and the list feature 10% of
> > the time.
>
> Wasn't your use case that you wanted to specify form fields in
> a given order (LIST), render a default view of the form in that
> order (LIST), and, later on, access the field specifiers in an
> arbitrary order, based on their key (DICT). Sure looks like it's
> the LIST aspect that's important here...
let's leave it there.
>
> > I want an ordered dict. Rather than a list and create this new view every
> > time and every where I want to use it as a dict.
>
> You want an ordered dict because you say you want one, not be-
> cause it's the best way to address your use case. That's fine, but
> it's not really related to the question asked in the subject line.
which is organized by YOU, that is a different story. As for related to
the subject line, since when I said my preference or use case has
anything to do with the subject line ? I have said in another post that
I don't think there should be one in the standard library, which is
directly about the subject line.
>
> > parsing or not parsing is not the point, and parsing/converting is
> > still "create a new view" of an existing data structure.
>
> Copying the entire data structure hardly qualifies as "creating a
> new view". dict() doesn't do that; in this use case, it doesn't cost
> you anything to use it.
doesn't cost me anything ? That is good news to me.
Antoon Pardon
> bon...@gmail.com wrote:
>> Personally, I have needs for ordered dict but I don't think it should
>> be in standard library though, as different situation called for
>> different behaviour for "ordered" and skewing my code to a standard lib
>> way is no good.
>
> I have started the thread in the first place because I believed it is
> pretty unabmiguous what an "ordered dictionary" is and how it should
> behave. That's why I asked myself why something that straigthforward has
> not been added to the standard lib yet. Maybe I'm wrong; I must admit
> that I haven't meditated about it very much.
Well it doesn't seem that obvious, because the two recipes you have
gotten, do something different from what I understand as an ordered
dictionary.
The two recipes order the keys by insertion order.
My idea would have been that some order was defined
on your keys in advance and that when you iterated over
the dictionary, the results would be ordered in sequence
of key order.
> Do you have an example for different options of behavior?
Well you have two above. Maybe someone can think of something
else.
Which behaviour are you looking for?
--
Antoon Pardon
Ben Sizer
> > as we can always convert a string field into some other types when it
> > is needed.
>
> No, that's not the same logic. The dict() in my example doesn't convert be-
> tween data types; it provides a new way to view an existing data structure.
This is interesting; I would have thought that the tuple is read and a
dictionary created by inserting each pair sequentially. Is this not the
case?
--
Ben Sizer
Fredrik Lundh
> > No, that's not the same logic. The dict() in my example doesn't convert be-
> > tween data types; it provides a new way to view an existing data structure.
>
> This is interesting; I would have thought that the tuple is read and a
> dictionary created by inserting each pair sequentially. Is this not the
> case?
pointers to the members of each pair, yes. but a pointer copy is a
cheap operation (for the given use case, we're only talking about a
few dozen pairs anyway, at the most).
this is a common fallacy; Python programmers underestimate the
cost of adding extra layers to their code (e.g. by using an ordered
dict data structure that has to incrementally update both a list and
a dictionary), and overestimate the cost of letting the C layer do
bulk operations.
(as an example, on my machine, using Foord's OrderedDict class
on Zwerschke's example, creating the dictionary in the first place
takes 5 times longer than the index approach, and accessing an
item takes 3 times longer. you can in fact recreate the index 6
times before OrderedDict is faster; if you keep the index around,
the OrderedDict approach never wins...)
</F>
Fuzzyman
Fredrik Lundh wrote:
> Ben Sizer wrote:
>
> > > No, that's not the same logic. The dict() in my example doesn't convert be-
> > > tween data types; it provides a new way to view an existing data structure.
> >
> > This is interesting; I would have thought that the tuple is read and a
> > dictionary created by inserting each pair sequentially. Is this not the
> > case?
>
> (as an example, on my machine, using Foord's OrderedDict class
> on Zwerschke's example, creating the dictionary in the first place
> takes 5 times longer than the index approach, and accessing an
> item takes 3 times longer. you can in fact recreate the index 6
> times before OrderedDict is faster; if you keep the index around,
> the OrderedDict approach never wins...)
>
So, so long as you want to use the dictionary less than six times -
it's faster to store/access it as a list of tuples. ;-)
Everytime you want to access (or assign to) the data structure as a
dictionary, you have to re-create the index.
Fuzzyman
http://www.voidspace.org.uk/python/index.shtml
> </F>
Alex Martelli
...
> d = somedict_from_db()
> prefer=['f','a',b']
>
> def my_order(d):
> for x in prefer:
> if x in d: yield x
> s = frozenset(prefer)
> for x in d:
> if x not in s: yield x
Yes, a much cleaner architecture (if you don't need any sorting on
non-preferred keys of d) than the ponderous general one I proposed. A
'key' approach with this behavior would be:
def my_key(k):
try: return prefer.index(k)
except ValueError: return len(prefer)
Now, 'for x in sorted(d, key=my_key)' should be equivalent to 'for x in
my_order(d)' thanks to the stability of sorting when the 'key' callable
returns equal values.
Neither of these way-simpler approaches is (I suspect) optimal for
speed, in the unlikely event one cares about that. The idea of
preprocessing the 'preferred' list once and for all outside of the
function (which I used heavily in my previous post) might yield some
speed-up, for example:
def my_key_fast(k, _aux=dict((k,i) for i,k in enumerate(prefer),
_l=len(prefer)):
return _aux.get(k, _l)
It's very unlikely that this situation warrants such optimization, of
course, I'm just "thinking aloud" about abstract possibilities.
Alex
Christoph Zwerschke
>>> Another possibility: ordered dictionaries are not needed when Python
>>> 2.4 has the 'sorted' builtin.
Christoph Zwerschke wrote:
>> The 'sorted' function does not help in the case I have indicated,
>> where "I do not want the keys to be sorted alphabetically, but
>> according to some criteria which cannot be derived from the keys
>> themselves."
Mike Meyer wrote:
> And how would an ordered dictionary help in this case?
Maybe there is some confusion between an "orderable" and an "ordered"
dictionary. When I talk about "ordered dictionary", then in the simplest
case I just set up my ordered dictionary with my preferred key order and
it stays like that. This allows me to later iterate through the
dictionary in this preferred order, while being still able to randomly
access data from the dictionary at other places.
-- Christoph
Alex Martelli
...
> ("but assume that I have some other use case" isn't a valid use
> case)
+1 QOTW
Alex
Christoph Zwerschke
> (as an example, on my machine, using Foord's OrderedDict class
> on Zwerschke's example, creating the dictionary in the first place
> takes 5 times longer than the index approach, and accessing an
> item takes 3 times longer. you can in fact recreate the index 6
> times before OrderedDict is faster; if you keep the index around,
> the OrderedDict approach never wins...)
You're right; I found creating a Larosa/Foord OrderedDict in this
example to be even 8 times slower than an ordinary dict. However, two
things need to be said here: 1) The dictionary in my exmaple was pretty
small (only 3 items), so you are not really measuring the performance of
the ordered dict, but mainly the overhead of using a user derived class
in comparison with the built-in dict type. And 2) the implementation by
Larosa/Foord is very slow and can be easily improved, for instance like
that:
def __init__(self, init_val = ()):
dict.__init__(self, init_val)
self.sequence = [x[0] for x in init_val]
With this change, creating the ordered dictionary is considerably faster
and for an average size dictionary, the factor of 8 pretty quickly
converges against 1.
But of course, it will always be slower since it is constructed on top
of the built-in dict. In end effect, you always have to maintain a
sequence *plus* a dictionary, which will be always slower than a sheer
dictionary. The ordered dictionary class just hides this uglyness of
having to maintain a dictionary plus a sequence, so it's rather an issue
of convenience in writing and reading programs than a performance issue.
It may be different if the ordered dict would be implemented directly as
an ordered hash table in C.
-- Christoph
Christoph Zwerschke
> Note the plural in 'insertion orderS': some people care about the FIRST
> time a key was added to a dict, some about the LAST time it was added,
> some about the latest time it was 'first inserted' (added and wasn't
> already there) as long as it's never been deleted since that occasion --
> and these are just a few of the multifarious orders based on the time of
> insertions and deletions of keys.
Ok, I start to understand that ambiguity emerges when you delete and
insert items. I didn't think much about this problem because my use
cases usually do not involve inserttion or deletion after the ordered
dictionary has been created.
But I think the following rule is "natural" enough to consider it as THE
standard behavior of ordered dictionaries:
"Insertion: If the key exists: Don't change the order. If it does not
exist: Append it to the sequence of keys. Deletion: Remove from the
sequence of keys."
I think this is also the behavior of associative arrays in PHP or Perl
and could be considered as the "ONE unambiguous definition".
-- Christoph
Aahz
Alex Martelli <al...@mail.comcast.net> wrote:
>
>I think you're wrong here. People in the past who have requested or
>implemented stuff they called 'ordered dicts' in the past had in mind
>drastically different things, based on some combination of insertion
>orders, keys, and _values_. So, ambiguity is definitely present in the
>phrase 'ordered dictionary', because there are so many different
>criteria whereby the 'ordering' could take place.
>
>Note the plural in 'insertion orderS': some people care about the FIRST
>time a key was added to a dict, some about the LAST time it was added,
>some about the latest time it was 'first inserted' (added and wasn't
>already there) as long as it's never been deleted since that occasion --
>and these are just a few of the multifarious orders based on the time of
>insertions and deletions of keys.
Ayup. In our application, not only do we have ordered dicts, we also
have something called a "sectioned dict", which is a dict-like object
that also looks like a regular class instance with attribute access. The
section part actually has multiple dicts (the sections) which are
layered, so that a dict key in the top layer overrides the value of the
key in lower layers. We traditionally have used it such that the
sections are accessed in MRU orders; last week, we added a new feature
that allows setting section values without changing section order (to
allow setting a default, essentially).
--
Aahz (aa...@pythoncraft.com) <*> http://www.pythoncraft.com/
"If you think it's expensive to hire a professional to do the job, wait
until you hire an amateur." --Red Adair
Tom Anderson
> Christoph Zwerschke <ci...@online.de> wrote:
>
>> The 'sorted' function does not help in the case I have indicated, where
>> "I do not want the keys to be sorted alphabetically, but according to
>> some criteria which cannot be derived from the keys themselves."
>
> Ah, but WHAT 'some criteria'? There's the rub! First insertion, last
> insertion, last insertion that wasn't subsequently deleted, last
> insertion that didn't change the corresponding value, or...???
All the requests for an ordered dictionary that i've seen on this group,
and all the cases where i've needed on myself, want one which behaves like
a list - order of first insertion, with no memory after deletion. Like the
Larosa-Foord ordered dict.
Incidentally, can we call that the "Larosa-Foord ordered mapping"? Then it
sounds like some kind of rocket science discrete mathematics stuff, which
(a) is cool and (b) will make Perl programmers feel even more inadequate
when faced with the towering intellectual might of Python. Them and their
Scwartzian transform. Bah!
tom
--
Baby got a masterplan. A foolproof masterplan.
Kay Schluehr
> huh? if you want a list, use a list.
>
> d = [('a', {...}), ('b', {....})]
If one wants uniform access to a nested data structure like this one
usually starts writing a wrapper class. I do not think the requirement
is anyhow deeper than a standard wrapper around such a list ( as a
model ) but the implementation may be different with respect to optimal
time complexitiy of element access. But the interface of the wrapper
class of d might resemble that of a dict. While the interface is that
of a dict the implementation is closer to a nested list. An "ordered
dict" would lower the impedance between a dict and a list.
Kay
Fredrik Lundh
> [snip..]
> > (as an example, on my machine, using Foord's OrderedDict class
> > on Zwerschke's example, creating the dictionary in the first place
> > takes 5 times longer than the index approach, and accessing an
> > item takes 3 times longer. you can in fact recreate the index 6
> > times before OrderedDict is faster; if you keep the index around,
> > the OrderedDict approach never wins...)
>
> So, so long as you want to use the dictionary less than six times -
> it's faster to store/access it as a list of tuples. ;-)
nope. that's not what I said. I said that you can recreate the index
six times in the time it takes to create a single OrderedDict instance.
if you need to use index more than that, it's not that hard to keep a
reference to it.
> Everytime you want to access (or assign to) the data structure as a
> dictionary, you have to re-create the index.
the use case we're talking about here (field descriptors) doesn't involve
assigning to the data structure, once it's created.
I'll repeat this one last time: for the use cases presented by Zwerschke
and "bonono", using a list as the master data structure, and creating the
dictionary on demand, is a lot faster than using a ready-made ordered
dict implementation. if you will access things via the dictionary a lot,
you can cache the dictionary somewhere. if not, you can recreate it
several times and still get a net win.
for other use cases, things may be different, but nobody has presented
such a use case yet. as I said earlier, "let's assume we have another use
case" is not a valid use case.
</F>
Christoph Zwerschke
> I'll repeat this one last time: for the use cases presented by Zwerschke
> and "bonono", using a list as the master data structure, and creating the
> dictionary on demand, is a lot faster than using a ready-made ordered
> dict implementation. if you will access things via the dictionary a lot,
> you can cache the dictionary somewhere. if not, you can recreate it
> several times and still get a net win.
You're right in pointing out that the advantage of ordered dictionaries
(unless you use an omptimized C implementation) is not a performance gain.
But please see my other reply: If the dictionary has more than 3 items
(say 10 or 20), and an effective ordered dict is used, it's not really
"a lot" slower. At least if we are talking about a situation were "on
demand" is "always". So, on the other side there isn't such a big
performance loss when using ordered dictionaries as well.
The advantage of using an ordered dictionary is that you can set up your
ordered dictionary (say, describing your database columns) once, and
then can access it in any way you like in the following: Iterate over it
in a guaranteed order or access item, always refering to the same
object, without needing to care about building and caching auxiliary
objects with different names depending on what you are doing.
-- Christoph
Bengt Richter
>
>Bengt Richter wrote:
>> On Sun, 20 Nov 2005 22:03:34 +0100, Christoph Zwerschke <ci...@online.de> wrote:
>> >> Ordering the keys isn't the normal case, and can be done easily when
>> >> needed.
>> >
>> >That depends. Maybe I do not want the keys to be sorted alphabetically,
>> >but according to some criteria which cannot be derived from the keys
>> >themselves.
>> You mean involving also the values? What's wrong with
>> sorted(plaindict.items(), key=your_ordering_function) ?
>>
>Not according to the content of the data, not just the "key". Or in
>other words, some other metadata that is not present in the data. A
>typical thing, like order of creation. Or some arbitary order. For
>example :
>
>I present a data grid/table in a HTML form and the user just drag and
>drop and rearrange the columns order.
^^[1]
[1] implies known info of before and after rearrangement. Where do these
come from, and are the two states expressed as ordered sets of keys generated and stored somewhere?
The point is, to re-order, you need a mapping from unordered data dict keys to values which the sorted
builtin function will order in the way you want. (BTW, if you use DSU, make sure the data is not modifying
your sort in an undesired way. Passing a key function to sorted makes it easy to exclude unwanted data from
the sort). If you have data that determines a new ordering of keys, it has to be accessed somehow, so
you just need to make it accessible to a handy helper that will generate your key function. E.g,
with before and after lists of keys expressing e.g. drag-drop before and after orderings, lambda can do the
job of getting you dict items in the new order, e.g., where bef and aft are lists that define the desired orderings
before and after in the sense of sort_precedence = bef.index[key_in_bef] and same for aft.
sorted(thedict.items(),key=lambda t:dict(zip(bef,((k in aft and aft.index(k) or len(aft)+bef.index(k)) for k in bef))[t[0]])
Ok, that one-liner grew a bit ;-)
>
>Of course, you may say, just put another column that represent
>this(some reporting programs I have seen do it this way) and that is an
>option but not the only option.
>
Maybe you could keep the rearranged_keys vector in a per-user cookie, if it's a web app
and amounts to a user personalization?
( posting delayed >12 hrs due to news server prob ;-/ )
Regards,
Bengt Richter
Bengt Richter
>Fredrik Lundh wrote:
>> if you restructure the list somewhat
>> d = (
>> ('pid', ('Employee ID', 'int')),
>> ('name', ('Employee name', 'varchar')),
>> ('sal', ('Salary', 'float'))
>> )
>> you can still loop over the list
>> ...
>> but you can easily generate an index when you need it:
>> index = dict(d)
>
>That's exactly the kind of things I find myself doing too often and what
>I was talking about: You are using *two* pretty redundant data
>structures, a dictionary and a list/tuple to describe the same thing.
>Ok, you can use a trick to automatically create the dictionary from the
>tuple, but still it feels somewhat "unnatural" for me. A "ordered
>dictionary" would be the more "natural" data structure here.
>
But, as has been mentioned**n, this is only one example of an ordering one
could make default for an "ordered" dictionary. Suppose you say it should
be ordered by insertion order, so
d = OrderedDict(); d[1]='one'; d[2]='two' =>> list(d) => [1, 2]
ok, now we do d[1]='ein' and what is the order? list(d) => [2, 1] ??
Or do replacements not count as "insertions"? The devil is always going
to be in the details. Maybe you want a model that works more like a list
of key:value pairs with just optimized access to a pair by key name as
well as position in the list. Or maybe you want to permit append and
NOT prevent [('a',1), ('a':2)] and maybe d['a'] => [1, 2] ???
The point is that Python is a nice lego set, and pre-molded castles
don't re-use well, even if they suit a particular you to a t ;-)
Note that is isn't hard to snap a few pieces together to make an ordered
dict to your own specs. But IMO it belongs in pyPI or such, not in the system
library. At least until it gets a lot of mileage -- and MMV ;-)
>I also wanted to mention the uglyness in the definition (nested tuples),
>but then I understood that even an ordered dictionary would not
>eliminate that uglyness, since the curly braces are part of the Python
>syntax and cannot be used for creating ordered dictionaries anyway. I
>would have to define the ordered dictionary in the very same ugly way:
>
>d = odict(('pid', ('Employee ID', 'int')),
> ('name', ('Employee name', 'varchar')),
> ('sal', ('Salary', 'float')))
>
>(Unless the Python syntax would be extend to use double curly braces or
>something for ordered dictionaries - but I understand that this is not
>an option.)
>
Whatever your odict does, if I had type a lot of definitions for it
I think I would write a QnD helper to make this work:
d = odict(prep("""
pid, Employee ID, int
name, Employee name, varchar # (comments to be ignored)
sal, Salary, float # alignment as above not mandatory
other, Something else, long, additional elements, allowed in second tuple?
"""))
Bengt Richter
>
>Fredrik Lundh wrote:
>> bon...@gmail.com wrote:
>>
>> > If I need the dict feature 90% of the time, and the list feature 10% of
>> > the time.
>>
>> Wasn't your use case that you wanted to specify form fields in
>> a given order (LIST), render a default view of the form in that
>> order (LIST), and, later on, access the field specifiers in an
>> arbitrary order, based on their key (DICT). Sure looks like it's
>> the LIST aspect that's important here...
>Yes. But whether LIST aspect or DICT is important is well, opinion. So
>let's leave it there.
>>
>> > I want an ordered dict. Rather than a list and create this new view every
>> > time and every where I want to use it as a dict.
>>
>> You want an ordered dict because you say you want one, not be-
>> cause it's the best way to address your use case. That's fine, but
>> it's not really related to the question asked in the subject line.
>Again, best way is decided by ME. If I am entering a coding contest
>which is organized by YOU, that is a different story. As for related to
>the subject line, since when I said my preference or use case has
>anything to do with the subject line ? I have said in another post that
>I don't think there should be one in the standard library, which is
>directly about the subject line.
Ok, so if not in the standard library, what is the problem? Can't find what
you want with google and PyPI etc.? Or haven't really settled on what your
_requirements_ are? That seems to be the primary problem people who complain
with "why no sprollificator mode?" questions. They don't know what they really
mean when it comes down to a DYFR (Define Your Felicitous Requirements) challenge.
So DYFR ;-)
Then someone can take less time than many of these posts takes to make a
list subclass that also acts like the dict when you want or a dict subclass that
also acts like a list when you want. Which methods from which would you like
as-is, and which modified? Any additional methods or properties? DYFR ;-)
>
>>
>> > parsing or not parsing is not the point, and parsing/converting is
>> > still "create a new view" of an existing data structure.
>>
So you'd like the mechanics to be automated and hidden? Then you need to
DYFR for using the black box you want. Methods, semantics.
>> Copying the entire data structure hardly qualifies as "creating a
>> new view". dict() doesn't do that; in this use case, it doesn't cost
>> you anything to use it.
>doesn't cost me anything ? That is good news to me.
Well, if you want something specific, it WILL cost you the effort to DYFR in detail ;-)
Alex Martelli
...
> But I think the following rule is "natural" enough to consider it as THE
> standard behavior of ordered dictionaries:
>
> "Insertion: If the key exists: Don't change the order. If it does not
> exist: Append it to the sequence of keys. Deletion: Remove from the
> sequence of keys."
>
> I think this is also the behavior of associative arrays in PHP or Perl
Perl hashes now keep track of 'order of keys'? That's new to me, they
sure didn't back when I used Perl! It's been a while, but a little
googling shows me, e.g at
<http://www.openarchives.org/pipermail/oai-implementers/2002-September/0
00642.html>, assertions such as:
"""
Hashes don't maintain key order. To get them in sorted order try:
foreach $i (sort keys(%afiliacao))
"""
which fully match my memories. Could you produce a URL to support the
hypothesis that Perl has changed its behavior? What about PHP? Thanks!
> and could be considered as the "ONE unambiguous definition".
"first insertion (since the last deletion if any)" is ONE unambiguous
definition, but surely not "_the_ ONE" with emphasis on ``the''. I see
nothing _ambiguous_ (nor _unnatural_) in being interested in the *last*
insertion, for example; indeed if phrased as "upon insertion, put the
key at the end of the sequence" (whether it was already elsewhere in the
sequence of not), with no need for conditionals regarding previous
existence, it might appear more "conceptually compact".
Anyway -- subclassing dict to implement your definition is reasonably
easy, and we could put the resulting package on the Cheese Shop. I hope
python.org keeps good enough statistics to be able to tell us, a couple
months later, how many people downloaded said package, vs how many
people downloaded a complete Python distro; of course, that ratio is
biased (in favour of the package) by the fact that many people already
have a complete distro available, while initially nobody would have the
package, but if we measure when things settle, after letting a month of
two or 'transient' pass, that effect might be lessened.
If we ran such an experiment, what fraction do you think would serve to
convince Guido that a dict 'ordered' by your definition is necessary in
Python 2.5's standard library (presumably in module 'collections')?
Alex
Fredrik Lundh
> Incidentally, can we call that the "Larosa-Foord ordered mapping"?
The implementation, sure.
> Then it sounds like some kind of rocket science discrete mathematics stuff
But math folks usually name things after the person(s) who came
up with the idea, not just some random implementer. The idea of
combining unordered mappings and ordered sequences is older than
Python itself.
</F>
Christoph Zwerschke
> Perl hashes now keep track of 'order of keys'? That's new to me, they
> sure didn't back when I used Perl!
Maybe I shouldn't have talked about Perl when I'm an ignoramus about
that language... You're right, Perl has unordered arrays. That was new
to me since I associate the term "array" always with "ordered" and I
just remembered that PHP's assoc arrays are similar to Perl's but in
fact, and the examples I have read did not mention about that problem.
> What about PHP?
You can conclude that PHP's assoc arrays are ordered from the fact that
the language provides a ksort() function to order the keys. And I think
PHP's insertion order is the one I mentioned in my last post.
Anyway, it would be interesting to examine this in detail and how this
is implemented in other languages.
> "first insertion (since the last deletion if any)" is ONE unambiguous
> definition, but surely not "_the_ ONE" with emphasis on ``the''.
> I see nothing _ambiguous_ (nor _unnatural_) in being interested in the
> *last* insertion, for example; indeed if phrased as "upon insertion, put
> the key at the end of the sequence" (whether it was already elsewhere in
> the sequence of not), with no need for conditionals regarding previous
> existence, it might appear more "conceptually compact".
But it would not satisfy the concept of "keys of a dictionary" which are
always unique.
BTW, there are some boundary conditions that should be fulfilled for the
insertion order, most obviously:
If you define an ordered dict that way:
d = odict()
d['a'] = 1
d['b'] = 2
d['c'] = 3
The keys should then be orderes as ('a', 'b', 'c').
> Anyway -- subclassing dict to implement your definition is reasonably
> easy, and we could put the resulting package on the Cheese Shop. I hope
> python.org keeps good enough statistics to be able to tell us, a couple
> months later, how many people downloaded said package, vs how many
> people downloaded a complete Python distro; of course, that ratio is
> biased (in favour of the package) by the fact that many people already
> have a complete distro available, while initially nobody would have the
> package, but if we measure when things settle, after letting a month of
> two or 'transient' pass, that effect might be lessened.
That would be also biased (in favour of Python) by the fact that
probably very little people would look for and use the package in the
cheese shop if they were looking for ordered dicts. I for example would
probably use ordered dicts if they were part of the standard lib, but
not if I have to install it as an obscure separate package. So I don't
think that will give us a real clue how many people would like ordered
dicts in the standard lib.
But anyway, if I find some time, I will research a little bit more about
the issue and create such a package, because it seems to me that the
existing packages and recipes are not really satisfying and you're right
it seems to be reasonably easy. It's on my todo list now...
-- Christoph
Christoph Zwerschke
> Ok, so if not in the standard library, what is the problem? Can't find what
> you want with google and PyPI etc.? Or haven't really settled on what your
> _requirements_ are? That seems to be the primary problem people who complain
> with "why no sprollificator mode?" questions.
What I don't understand is why legitimate questions such as "why are
there no ordered dictionaries" are immediately interpreted as
*complaints* and not just as questions. If I ask such a question, I am
not complaining but trying to simply figure out *why* there is no such
thing. Probably there are reasons and all I want to know is find these
reasons and learn a little bit more about Python in doing so.
Why can't such questions be discussed in a factual, calm and friendly way?
> They don't know what they really mean when it comes down to a DYFR
> (Define Your Felicitous Requirements) challenge.
I don't think that this was true in this case, and even if this is the
outcome, those who asked the question will have learned something.
I think a discussion group is not there for only presenting mature,
sophisticated thoughts and concepts, but also for "thinking loud"
together with other about these issues. We all know that clarifying our
thoughts works often best if you discuss them with others. And I think
that's one purpose of discussion lists. Asking questions should not be
immediately be discouraged, even silly questions. If it is really a FAQ,
you can simply point to the FAQ or add the answer in the FAQ list if it
is missing there.
-- Chris
Fredrik Lundh
> What about PHP? Thanks!
according to some random PHP documentation I found on the intarweb:
An array in PHP is actually an ordered map. A map is a type that
maps values to keys.
and later:
A key may be either an integer or a string. If a key is the standard
representation of an integer, it will be interpreted as such (i.e. "8"
will be interpreted as 8, while "08" will be interpreted as "08"). Floats
in key are truncated to integer.
and later:
You cannot use arrays or objects as keys. Doing so will result in a
warning: Illegal offset type.
at which point my brain raised an exception.
</F>
Fuzzyman
> Fredrik Lundh wrote:
[snip..]
> You're right; I found creating a Larosa/Foord OrderedDict in this
> example to be even 8 times slower than an ordinary dict. However, two
> things need to be said here: 1) The dictionary in my exmaple was pretty
> small (only 3 items), so you are not really measuring the performance of
> the ordered dict, but mainly the overhead of using a user derived class
> in comparison with the built-in dict type. And 2) the implementation by
> Larosa/Foord is very slow and can be easily improved, for instance like
> that:
>
> def __init__(self, init_val = ()):
> dict.__init__(self, init_val)
> self.sequence = [x[0] for x in init_val]
>
But that doesn't allow you to create an ordered dict from another
ordered dict.
It also allows duplicates in the sequence attribute. It's a useful idea
though.
Thanks
Fuzzyman
http://www.voidspace.org.uk/python/index.shtml
Christoph Zwerschke
> d = OrderedDict(); d[1]='one'; d[2]='two' =>> list(d) => [1, 2]
> ok, now we do d[1]='ein' and what is the order? list(d) => [2, 1] ??
> Or do replacements not count as "insertions"?
If you simply set a value for a key that already exists, the order
should not be changed. I think this is intuitive.
> Or maybe you want to permit append and NOT prevent
> [('a',1), ('a':2)] and maybe d['a'] => [1, 2] ???
You could ask the same question about dict. I think that is not an
option. Why should you want odict behave different than dict?
I still believe that the concept of an "ordered dictionary" ("behave
like dict, only keep the order of the keys") is intuitive and doesn't
give you so much scope for ambiguity. But probably I need to work on an
implementation to become more clear about possible hidden subtleties.
-- Christoph
Kay Schluehr
> That would be also biased (in favour of Python) by the fact that
> probably very little people would look for and use the package in the
> cheese shop if they were looking for ordered dicts.
Does anyone actually use this site? While the Vaults offered a nice
place and a nice interface the Cheese Shop has the appeal of a code
slum.
Kay
Fuzzyman
Hmmm.. it's *the* repository for Python code. I expect quite a few
people use it...
:-)
Fuzzyman
http://www.voidspace.org.uk/python/index.shtml
> Kay
Steven D'Aprano
> Tom Anderson wrote:
>
>> Incidentally, can we call that the "Larosa-Foord ordered mapping"?
>
> The implementation, sure.
>
>> Then it sounds like some kind of rocket science discrete mathematics stuff
>
> But math folks usually name things after the person(s) who came
> up with the idea, not just some random implementer.
No no no! In maths things are usually named after Euler, or the first
person to discover them after Euler.
--
Steven.
Magnus Lycka
> But please see my other reply: If the dictionary has more than 3 items
> (say 10 or 20), and an effective ordered dict is used, it's not really
> "a lot" slower. At least if we are talking about a situation were "on
> demand" is "always". So, on the other side there isn't such a big
> performance loss when using ordered dictionaries as well.
There is no performance issue involved with this usecase at all!
It doesn't matter if it's hundreds of tuples of strings in a list
if it's supposed to be presented in a GUI. Recreating a dict from
that is bound to be magnitudes faster than getting the stuff
visible on the screen, at least if we're talking web apps. So is
using a reasonable odict implementation, if that's what you want.
I think the issue is not to overload the already extensive standard
library with trivial things that can easily be replaced by your own
three line wrapper, especially if there are a number of different
semantics that could be imagined for such a thingie.
The C++ std lib has an ordered "dict" class called map, and that's
ordered by key. Others suggested ordering by value, and there are a
number of different interpretations of the 'order by insertion time'
theme in the air. This clearly smells like "fix your own thing and
leave it out of the standard library".
With one of these solutions picked as Python's "ordered dict", we'll
make things slightly more convenient for a few programmers and just
burden others with something that sounds right for them, but turns
out not to solve their problems. Or, we could try to make a bunch
of different solution, increasing the cognitive burden for all who
learn Python while solving non-problems. This just isn't going to
happen!
bon...@gmail.com
Bengt Richter wrote:
> Ok, so if not in the standard library, what is the problem? Can't find what
> you want with google and PyPI etc.? Or haven't really settled on what your
> _requirements_ are? That seems to be the primary problem people who complain
> with "why no sprollificator mode?" questions. They don't know what they really
> mean when it comes down to a DYFR (Define Your Felicitous Requirements) challenge.
> So DYFR ;-)
> >> > parsing or not parsing is not the point, and parsing/converting is
> >> > still "create a new view" of an existing data structure.
> >>
> So you'd like the mechanics to be automated and hidden? Then you need to
> DYFR for using the black box you want. Methods, semantics.
Lose you. don't know what you want to say.
bon...@gmail.com
Christoph Zwerschke wrote:
> Bengt Richter schrieb:
> > Ok, so if not in the standard library, what is the problem? Can't find what
> > you want with google and PyPI etc.? Or haven't really settled on what your
> > _requirements_ are? That seems to be the primary problem people who complain
> > with "why no sprollificator mode?" questions.
>
> What I don't understand is why legitimate questions such as "why are
> there no ordered dictionaries" are immediately interpreted as
> *complaints* and not just as questions. If I ask such a question, I am
> not complaining but trying to simply figure out *why* there is no such
> thing. Probably there are reasons and all I want to know is find these
> reasons and learn a little bit more about Python in doing so.
>
> Why can't such questions be discussed in a factual, calm and friendly way?
>
Using "why can't" is already too much. Even you turn it into "is there
a thing for ordered dict", you would get similar treatment
Get used to it :-)
> > They don't know what they really mean when it comes down to a DYFR
> > (Define Your Felicitous Requirements) challenge.
>
> I don't think that this was true in this case, and even if this is the
> outcome, those who asked the question will have learned something.
>
> I think a discussion group is not there for only presenting mature,
> sophisticated thoughts and concepts, but also for "thinking loud"
> together with other about these issues. We all know that clarifying our
> thoughts works often best if you discuss them with others. And I think
> that's one purpose of discussion lists. Asking questions should not be
> immediately be discouraged, even silly questions. If it is really a FAQ,
> you can simply point to the FAQ or add the answer in the FAQ list if it
> is missing there.
Well, different groups has different "personality", just don't be
scared.
bon...@gmail.com
but how best to blend with the rest of the code which I have no
interest to explain as I am not here to convincing anyone for such a
thing. I just present a use case, if they see it, fine. If not, that is
fine too.
But I did learn something that creating a dict on a list cost me
nothing, I would be less worry about the performance hit in the future.
A.M. Kuchling
> Does anyone actually use this site? While the Vaults offered a nice
> place and a nice interface the Cheese Shop has the appeal of a code
> slum.
Looking at the Cheese Shop's home page at
http://cheeseshop.python.org/pypi, which lists recent package updates,
three packages were updated on 11/22, and three on 11/21. Two on
11/20, six on 11/19 (someone was busy!).
Looking at the Vaults's 'latest' page at
http://py.vaults.ca/apyllo.py?a=l, two packages were updated on 08/23,
and five on 08/06.
What would improve the Cheese Shop's interface for you?
--amk
Kay Schluehr
>From the treasures of the vaults to cheese shops - an upside down
evolution :( Personally I find the usability of the Vaults quite well
and it's design o.k. I guess it is a technical detail that causes the
Vaults to be phased out.
If I'd try something more innovative by myself I'd think about
accessing and filtering packages through the Python console itself.
Probably an enhanced standard console enabling SQL commands. Updates of
a local database ( e.g. SQLite ) should be cheap and may be performed
by only one request. Searching, filtering, loading, installing would be
done in a Python+SQL environment.
Kay
Fuzzyman
the sequence attribute to change the ordering.
I think we're looking at improving performance based on some of the
suggestions here - as well as possibly widening it to include some of
the alternative use cases. (Or at least Nicola Larosa will do it and
I'll criticise it).
All for another day though...
Alex Martelli
> Alex Martelli schrieb:
> > Perl hashes now keep track of 'order of keys'? That's new to me, they
> > sure didn't back when I used Perl!
>
> Maybe I shouldn't have talked about Perl when I'm an ignoramus about
> that language... You're right, Perl has unordered arrays. That was new
> to me since I associate the term "array" always with "ordered" and I
> just remembered that PHP's assoc arrays are similar to Perl's but in
> fact, and the examples I have read did not mention about that problem.
Perl's _arrays_ are a bit like Python _lists_, and ordered; it's the
_hashes_ that are a bit like Python _dicts_, and unordered. PHP's use
of "array" for both concepts may indeed be a bit confusing.
> > What about PHP?
>
> You can conclude that PHP's assoc arrays are ordered from the fact that
> the language provides a ksort() function to order the keys. And I think
> PHP's insertion order is the one I mentioned in my last post.
>
> Anyway, it would be interesting to examine this in detail and how this
> is implemented in other languages.
Yep. After just a bit of research I suspect you're right re PHP but
haven't found a specific reference-manual page URL about it.
> > "first insertion (since the last deletion if any)" is ONE unambiguous
> > definition, but surely not "_the_ ONE" with emphasis on ``the''.
> > I see nothing _ambiguous_ (nor _unnatural_) in being interested in the
> > *last* insertion, for example; indeed if phrased as "upon insertion, put
> > the key at the end of the sequence" (whether it was already elsewhere in
> > the sequence of not), with no need for conditionals regarding previous
> > existence, it might appear more "conceptually compact".
>
> But it would not satisfy the concept of "keys of a dictionary" which are
> always unique.
Why not? Since keys are unique, any 'sequence' of keys is a permutation
of a set, a perfectly well-defined concept.
> BTW, there are some boundary conditions that should be fulfilled for the
> insertion order, most obviously:
>
> If you define an ordered dict that way:
>
> d = odict()
> d['a'] = 1
> d['b'] = 2
> d['c'] = 3
>
> The keys should then be orderes as ('a', 'b', 'c').
Yep, but both 'first-insertion' and 'latest-insertion' (and many other
rules) meet that constraint.
> That would be also biased (in favour of Python) by the fact that
> probably very little people would look for and use the package in the
> cheese shop if they were looking for ordered dicts. I for example would
> probably use ordered dicts if they were part of the standard lib, but
> not if I have to install it as an obscure separate package. So I don't
> think that will give us a real clue how many people would like ordered
> dicts in the standard lib.
A package on cheese shop need not be "obscure" -- that depends on
announcing it well, etc. And the standard library is so huge that many
things inside IT are in fact obscure -- I find myself often pointing out
standard-library solutions to rather experienced Pythonistas who had
been rolling their own, unaware of the stdlib alternative (thus making
the stdlib even bigger has a cost...). So, I don't think this effect
invalidates the experiment; while not all who download an add-on would
like it in the stdlib, and vice versa, there is surely a correlation
between amount of interest/need for the add-on's functionality, and both
rate of downloads as well as interest in having it in the stdlib.
> But anyway, if I find some time, I will research a little bit more about
> the issue and create such a package, because it seems to me that the
> existing packages and recipes are not really satisfying and you're right
> it seems to be reasonably easy. It's on my todo list now...
It presumably should have a C implementation for speed and a Python one
for wide availability and ease of installation (the latter gives all the
advantages of prototyping in fixing the API &c, and is made especially
easy by the existence of UserDict.DictMixin in the stdlib). I'll be
glad to help out with the C part when it gets to that, just holler...
Alex
Alex Martelli
...
> But math folks usually name things after the person(s) who came
> up with the idea, not just some random implementer. The idea of
Wrong: you're forgetting Stigler's Law of Misonomy (which I imagine must
have NOT been discovered by Stigler...;-). The Poisson distribution was
in fact described earlier by Bernoulli, Gosset's z-test is universally
known as Student's t-test, etc, etc.
Salsburg's delightful "The Lady Tasting Tea" has a lot of fun with
Stigler's law in the footnotes...;-)
Alex
Magnus Lycka
> I still believe that the concept of an "ordered dictionary" ("behave
> like dict, only keep the order of the keys") is intuitive and doesn't
> give you so much scope for ambiguity.
Sure. Others think so too. The problem is that if you and these other
people actually write down exactly how this unambigous ordered dict
should behave, you'll end up with a dozen different sets of semantics,
which can be divided in at least three main groups.
People use different idioms, and often gather collections of classes
and functions that fit their style of coding and their typical problem
domains. There is nothing wrong with that, and they might well be made
publically available if they are more generally useful.
When adding things to the standard library, there are some more things
to consider, particularly for something general such as an odict class:
Is is general enough, and does it add enough value to outweigh the
increased cognitive burden of more classes/functions/types in the
library?
For a beginner, it's easy to believe that things would be easier if
the tool you use had already solved the problem you have at hand for
you, but it turns out that the world has so many problems, that you
would end up with a impenetrable forest of standard library modules
if we actually tried to achieve this.
Stuart McGraw
"A.M. Kuchling" <a...@amk.ca> wrote in message news:9smdnVdh0oN...@speakeasy.net...
[...]
> What would improve the Cheese Shop's interface for you?
Getting rid of those damn top level links to old versions.
Seeing a long list of old versions, when 99% of visitors are
only interested in the current version, is just visual noise,
and really lame. Move the old version links onto the page
describing the software.
Anton Vredegoor
> But of course, it will always be slower since it is constructed on top
> of the built-in dict. In end effect, you always have to maintain a
> sequence *plus* a dictionary, which will be always slower than a sheer
> dictionary. The ordered dictionary class just hides this uglyness of
> having to maintain a dictionary plus a sequence, so it's rather an issue
> of convenience in writing and reading programs than a performance issue.
>
> It may be different if the ordered dict would be implemented directly as
> an ordered hash table in C.
The problem with hashing is that one is storing data from a possibly
wildly varying range in a set of values with a limited range. That's
where the ordering problems come from in the first place. If one wants
to solve it once and for all one has to give up the speed advantage
that makes hashing so popular. I wonder if optimized C code using
bisect would be very much slower for small ranges?
The current set implementation uses dicts to implement sets, while sets
are a more basic data type than dicts. At least dicts and sets should
be derived from the same type. Just speaking from an intuitive point of
view of course :-)
Here's a sorted dict implementation without using hashes, which maybe
would be fast if implemented in C. The insertion order can be retrieved
using the keys and values lists from the object itself, items() gives a
sorted sequence.
Anton.
NB warning untested code.
When using mutables as keys which is possible by this implementation,
don't change the keys after they're used in the object.
from array import array
class VDict:
def __init__(self,sequence = []):
self.keys = []
self.values = []
self.unranks = array('L',[])
for key,value in sequence:
self[key] = value
def __setitem__(self,key,value):
keys = self.keys
values = self.values
unranks = self.unranks
n = len(keys)
i = self.bisect_left(key)
if i == n or keys[unranks[i]] != key:
keys.append(key)
values.append(value)
unranks.insert(i,n)
else:
values[i] = value
def __getitem__(self,key):
i = self.bisect_left(key)
return self.values[self.unranks[i]]
def bisect_left(self,key, lo=0, hi=None):
keys = self.keys
unranks = self.unranks
if hi is None:
hi = len(keys)
while lo < hi:
mid = (lo+hi)//2
if keys[unranks[mid]] < key:
lo = mid+1
else:
hi = mid
return lo
def __contains__(self,key):
keys = self.keys
unranks = self.unranks
n = len(keys)
i = self.bisect_left(key)
return i < n and keys[unranks[i]] == key
def __len__(self):
return len(self.keys)
def items(self):
keys = self.keys
values = self.values
unranks = self.unranks
return [(keys[i],values[i]) for i in unranks]
def __iter__(self):
return iter(self.items())
def remove(self,key):
keys = self.keys
values = self.values
unranks = self.unranks
n = len(keys)
i = self.bisect_left(key)
x = unranks[i]
if i < n and keys[x] == key:
del keys[x]
del values[x]
del unranks[i]
for j,k in enumerate(unranks):
if k > x:
unranks[j] -= 1
def test():
V = VDict()
L = [1,2,3]
V[L] = 10
print V[L]
V[L] = 20
print V[L]
V.remove(L)
print V.items()
V = VDict(zip('edcba',range(5)))
print V.items()
print V['d']
V.remove('d')
print V.items()
if __name__=='__main__':
test()
Ben Sizer
> Ben Sizer wrote:
> > This is interesting; I would have thought that the tuple is read and a
> > dictionary created by inserting each pair sequentially. Is this not the
> > case?
>
> pointers to the members of each pair, yes. but a pointer copy is a
> cheap operation (for the given use case, we're only talking about a
> few dozen pairs anyway, at the most).
I was really thinking more about the other work, such as the hashing
and whatever, but I guess that is very efficient anyway.
> this is a common fallacy; Python programmers underestimate the
> cost of adding extra layers to their code (e.g. by using an ordered
> dict data structure that has to incrementally update both a list and
> a dictionary), and overestimate the cost of letting the C layer do
> bulk operations.
If it was me I would probably have just used a list and searched it
linearly: premature optimisation is the root of all evil, etc. But then
I've never found a need for an ordered dictionary anyway; I always felt
they were more an artifact of the language implementation than a
reflection of something inherently useful.
However, you have to forgive people for falling prey to the 'fallacy'
you describe - for years there's been an attempt to teach people to use
proper data structures and algorithms instead of relying on
micro-optimisations (ie. "it's too slow: redo it in assembly"). So
often, the first port of call for a good programmer will be to try and
find a structure that maps directly to the problem.
--
Ben Sizer
Fredrik Lundh
> > What would improve the Cheese Shop's interface for you?
>
> Getting rid of those damn top level links to old versions.
> Seeing a long list of old versions, when 99% of visitors are
> only interested in the current version, is just visual noise,
> and really lame. Move the old version links onto the page
> describing the software.
hmm? the pypi package automatically hides old versions when
you post new ones, and it's been that way for ages...
(which is bloody annoying if you're a package developers, since it
means that alphas for the next release hides the most recent stable
version)
looking at the full index, ZODB seems to be the only package that's
available in more than just one stable and one development version...
</F>
Bengt Richter
>
>Kay Schluehr wrote:
>> Christoph Zwerschke wrote:
>>
>> > That would be also biased (in favour of Python) by the fact that
>> > probably very little people would look for and use the package in the
>> > cheese shop if they were looking for ordered dicts.
>>
>> Does anyone actually use this site? While the Vaults offered a nice
>> place and a nice interface the Cheese Shop has the appeal of a code
>> slum.
>>
>
>Hmmm.. it's *the* repository for Python code. I expect quite a few
>people use it...
>
>:-)
>
I hadn't realized how much stuff was there. I generally google for stuff,
but I will be looking directly now.
BTW, I made a mod[1] to your odict that I think/hope is going to be generally faster.
It requires 2.4 though. It passes the same doctest, so its probably close to the same.
It stores the ordering info differently, but is also a dict subclass.
Do you happen to have a timing test that exercises various aspects, so I can try it
before I embarrass myself? Otherwise I guess I'll write something.
Would the would-be users chime in with some idea of what kinds of operations are
most important timing-wise? Which would get the most use? How dynamic would ordering typically be?
[1] fairly pervasive little mods actually
[ 9:15] C:\pywk\clp>diff odict.py odictb.py |wc
146 458 4948
[ 9:15] C:\pywk\clp>wc odict*.py
467 1228 12483 odict.py
511 1500 14728 odictb.py
978 2728 27211 Totals
Regards,
Bengt Richter
Stuart McGraw
http://cheeseshop.python.org/pypi?:action=browse&asdf=405
- ClientForm-0.1.17
- ClientForm-0.2.1b
...
- EmPy_3.1
- EmPy_3.1.1
- EmPy_3.2
- EmPy_3.3
...
- FauxIdent-1.1
- FauxIdent-1.2
- FauxIdent-1.2.1
...
Well, it is better than I remember it being a while (year?) ago, my
recollection is that many packages had many, many old versions
listed but now I usualy see only a couple versions.
Hmm, so two versions means one is a development version,
and the other is a stable version? I did not know that, and did
not see it documented on the site. I would say documenting
that would be an interface improvement.
I still think it would be better to have just a package name
(with current version) listed in the index page(s), and have alternate
versions (old, alpha testing, etc) listed on the package's description
page.
Bengt Richter
>Bengt Richter schrieb:
>> Ok, so if not in the standard library, what is the problem? Can't find what
>> you want with google and PyPI etc.? Or haven't really settled on what your
>> _requirements_ are? That seems to be the primary problem people who complain
>> with "why no sprollificator mode?" questions.
>
>What I don't understand is why legitimate questions such as "why are
>there no ordered dictionaries" are immediately interpreted as
>*complaints* and not just as questions. If I ask such a question, I am
>not complaining but trying to simply figure out *why* there is no such
>thing. Probably there are reasons and all I want to know is find these
>reasons and learn a little bit more about Python in doing so.
>
>Why can't such questions be discussed in a factual, calm and friendly way?
Sorry, I was tired and vented some impatience. I'm mostly friendly ;-)
I took the "why" in a different sense than it was meant, I guess.
Sort of like hearing "why haven't I been served yet" in a restaurant, and
having to say, "I don't work here, you'll have to ask the waiter."
>
> > They don't know what they really mean when it comes down to a DYFR
> > (Define Your Felicitous Requirements) challenge.
>
>I don't think that this was true in this case, and even if this is the
>outcome, those who asked the question will have learned something.
I agree again.
>
>I think a discussion group is not there for only presenting mature,
>sophisticated thoughts and concepts, but also for "thinking loud"
>together with other about these issues. We all know that clarifying our
>thoughts works often best if you discuss them with others. And I think
>that's one purpose of discussion lists. Asking questions should not be
>immediately be discouraged, even silly questions. If it is really a FAQ,
>you can simply point to the FAQ or add the answer in the FAQ list if it
>is missing there.
Agreed again. Thank you for your nice reply.
Regards,
Bengt Richter
Fredrik Lundh
> Hmm, so two versions means one is a development version,
> and the other is a stable version? I did not know that, and did
> not see it documented on the site. I would say documenting
> that would be an interface improvement.
well, that's up to the developer. when you upload a new version, all
older ones are automagically hidden. the only way to make old versions
appear again is to "unhide" them via a web form. for the few packages
I sampled, the older versions were stable, and the latest one was "less
stable", but I didn't check all of the...
> I still think it would be better to have just a package name
> (with current version) listed in the index page(s), and have alternate
> versions (old, alpha testing, etc) listed on the package's description
> page.
agreed.
a "nonstable"-property in the setup file would be nice too (so that stable
versions don't disappear when you upload an alpha or beta...)
</F>
Bengt Richter
>
>Bengt Richter wrote:
>> Ok, so if not in the standard library, what is the problem? Can't find what
>> you want with google and PyPI etc.? Or haven't really settled on what your
>> _requirements_ are? That seems to be the primary problem people who complain
>> with "why no sprollificator mode?" questions. They don't know what they really
>> mean when it comes down to a DYFR (Define Your Felicitous Requirements) challenge.
>> So DYFR ;-)
>Beat me. I am not the one asking the question.
Sorry, I thought you wanted an ordered dict too.
>
>> >> > parsing or not parsing is not the point, and parsing/converting is
>> >> > still "create a new view" of an existing data structure.
>> >>
>> So you'd like the mechanics to be automated and hidden? Then you need to
>> DYFR for using the black box you want. Methods, semantics.
>Lose you. don't know what you want to say.
>
I like solving problems. I just get frustrated when people don't focus on getting
the problem defined, which IME is 2/3 of the way to a solution. I don't mind,
in fact enjoy, rambling musings, but if someone seems actually to want a solution
for something, I like to try to realize it concretely.
After finally reading that the odict.py in PyPI by Larosa/Foord was what was desired,
but slower in use than what Fredrik posted, I decided to see if I could speed up odict.py.
I now have a version that I think may be generally faster.
I still don't know whether it will be of any user w.r.t. the requirements of anyone
on the bandwagon of asking for some kind of ordered dict, but we'll see what we'll see ;-)
Regards,
Bengt Richter
Bengt Richter
>Bengt Richter wrote:
>
> > d = OrderedDict(); d[1]='one'; d[2]='two' =>> list(d) => [1, 2]
> > ok, now we do d[1]='ein' and what is the order? list(d) => [2, 1] ??
> > Or do replacements not count as "insertions"?
>
>If you simply set a value for a key that already exists, the order
>should not be changed. I think this is intuitive.
>
>> Or maybe you want to permit append and NOT prevent
> > [('a',1), ('a':2)] and maybe d['a'] => [1, 2] ???
>
>You could ask the same question about dict. I think that is not an
>option. Why should you want odict behave different than dict?
Well, it was beginning to remind of RDB with possible non-unique keys,
where a select can get you multiple records back.
>
>I still believe that the concept of an "ordered dictionary" ("behave
>like dict, only keep the order of the keys") is intuitive and doesn't
>give you so much scope for ambiguity. But probably I need to work on an
>implementation to become more clear about possible hidden subtleties.
>
Does that mean that the Larosa/Foord odict.py implementation in PyPI
does not do what you want?
Regards,
Bengt Richter
Christoph Zwerschke
> Perl's _arrays_ are a bit like Python _lists_, and ordered; it's the
> _hashes_ that are a bit like Python _dicts_, and unordered. PHP's use
> of "array" for both concepts may indeed be a bit confusing.
Perl's _hashes_ have been also called _associative arrays_ originally.
>>Anyway, it would be interesting to examine this in detail and how this
>>is implemented in other languages.
Ok, I just did a little research an compared support for ordered dicts
in some other languages:
* Perl has hashes (associative arrays) which are not ordered.
Here also people are asking for and implementing "ordered hashes",
e.g. http://perltraining.com.au/tips/2005-06-29.html
http://search.cpan.org/dist/Tie-IxHash/lib/Tie/IxHash.pm
http://search.cpan.org/dist/Tie-InsertOrderHash/InsertOrderHash.pm
http://www.yapc.org/America/previous-years/19100/schedule/author/pinyan.html
* Ruby hashes are not ordered.
Again people are asking for and implementing "ordered hashes".
e.g. http://raa.ruby-lang.org/project/orderedhash/
http://groups.google.com/group/comp.lang.ruby/browse_frm/thread/8ebe8d1c5830c801/6428a870925f23f4
* Smalltalk: Innately has unordered Dictionaries.
People are asking for and implementing "OrderedDictionaries" as well,
e.g. http://exept.eu.org:8080/ClassDoc/classDocOf:,OrderedDictionary
* Lisp has (ordered) "association lists".
* PHP has ordered associative arrays.
* Awk, TCL: Associative arrays are unordered.
* C++ has a Map template in the STL which is ordered (a "Sorted
Associative Container").
* Java: Has "LinkedHashMap" which is ordered.
So ordered dictionaries don't seem to be such an exotic idea.
All implementations I found were pretty unequivocally the same that I
had in mind, using insertion order, appending the latest inserted keys
at the end, not changing the order if an existing key is re-inserted,
and deleting keys together with entries.
I also found a discussion thread like the current where similar
arguments were mentioned for and against ordered dictionaries:
In http://mail.python.org/pipermail/python-dev/2005-March/052041.html,
Nick Coghlan raises the following rhetorical question:
Would the default semantics below really be that suprising?
"An ordered dictionary remembers the order in which keys are first seen
and, when iterated over, returns entries based on that order. This
applies to direct iteration, iteration over values and (key, value)
pairs, to the list-producing methods (i.e. keys(), values() and items())
and to any other operations that involve implicit iteration (e.g.
converting to a string representation). Overwriting an entry replaces
its value, but does not affect its position in the key order. Removing
an entry (using 'del') _does_ remove it from the key order. Accordingly,
if the entry is later recreated, it will then occur last in the key
order. This behaviour is analagous to that of a list constructed using
only list.append() to add items (indeed, the key order can be thought of
as a list constructed in this fashion, with keys appended to the list
when they are first encountered)."
This was also the semantics I immediately had in mind when I thought
about ordered dictionaries, though I hadn't read anything about ordered
dictionaries before and it is the same semantics that I found others
have implemented in other languages.
I can't help but I still find it unambiguous and intuitive enough to
consider it "the one" standard implementation for ordered dictionaries.
Also, in the use cases mentioned (describing database columns, html form
fields, configuration parameters etc.), the dictionary is usually only
created once and then not changed, so different handling of re-insertion
or deletion of keys would not even be relevant in these cases.
-- Christoph
Christoph Zwerschke
> Of course ours is ordered *and* orderable ! You can explicitly alter
> the sequence attribute to change the ordering.
What I actually wanted to say is that there may be a confusion between a
"sorted dictionary" (one where the keys are automatically sorted) and an
"ordered dictionary" (where the keys are not automatically ordered, but
have a certain order that is preserved). Those who suggested that the
"sorted" function would be helpful probably thought of a "sorted
dictionary" rather than an "ordered dictionary."
-- Christoph
Christoph Zwerschke
> I'm mostly friendly ;-)
I'm pretty sure you are :-)
-- Chris
Christoph Zwerschke
> After finally reading that the odict.py in PyPI by Larosa/Foord was what was desired,
> but slower in use than what Fredrik posted, I decided to see if I could speed up odict.py.
> I now have a version that I think may be generally faster.
Hm, I wouldn't formulate it that way that I want the odict of
Larosa/Foord, but I want "the one" "obvious" odict for which
Larosa/Foord have already made one implementatin ;-)
Others are here:
http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/438823
http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/107747
http://pleac.sourceforge.net/pleac_python/hashes.html#AEN250
It is all essentially the same idea I think (though after having a
closer look I see implementation shortcomings in all of them).
> I now have a version that I think may be generally faster.
Great. I also wanted to do that. Also, I would like to add some
functionality to Larosa/Foord's odict, like creating or updating an
odict from an ordinary dict (keys taken over from the ordinary dict will
be either in random order or automatically sorted). An ordered
dictionary should also have methods for sorting (like PHP's ksort()).
This way, you could initialize an ordered dict from an ordinary dict,
sort it, and from then on never care to call keys().sorted() or
something when iterating over the dictionary. Probably there are other
methods from lists that could be taken over to ordered dicts.
-- Christoph
Christoph Zwerschke
>> dict.__init__(self, init_val)
>> self.sequence = [x[0] for x in init_val]
Fuzzyman wrote:
> But that doesn't allow you to create an ordered dict from another
> ordered dict.
Right; I did not want to present a full implementation, just the
relevant part. Of course, a real implementation should also allow to
build an ordered dict from another ordered dict or an ordinary dict. (In
the latter case, maybe the keys should be automatically sorted.) But one
or two case disctinctions would not make things mentionable slower.
-- Christoph
Carsten Haese
> Would the default semantics below really be that suprising?
>
> "An ordered dictionary remembers the order in which keys are first seen
> an entry (using 'del') _does_ remove it from the key order. Accordingly,
> if the entry is later recreated, it will then occur last in the key
> I can't help but I still find it unambiguous and intuitive enough to
> consider it "the one" standard implementation for ordered dictionaries.
I don't think it's intuitive if you can't describe it without
contradicting yourself. If the order of the keys really were the order
in which they were first seen by the dictionary, deleting and recreating
a key should maintain its original position.
-Carsten
Kay Schluehr
> On Mon, 21 Nov 2005 01:27:22 +0100, Christoph Zwerschke <ci...@online.de> wrote:
>
> >Fredrik Lundh wrote:
> >> if you restructure the list somewhat
> >> d = (
> >> ('pid', ('Employee ID', 'int')),
> >> ('name', ('Employee name', 'varchar')),
> >> ('sal', ('Salary', 'float'))
> >> )
> >> you can still loop over the list
> >> ...
> >> but you can easily generate an index when you need it:
> >> index = dict(d)
> >
> >That's exactly the kind of things I find myself doing too often and what
> >I was talking about: You are using *two* pretty redundant data
> >structures, a dictionary and a list/tuple to describe the same thing.
> >Ok, you can use a trick to automatically create the dictionary from the
> >tuple, but still it feels somewhat "unnatural" for me. A "ordered
> >dictionary" would be the more "natural" data structure here.
> >
> But, as has been mentioned**n, this is only one example of an ordering one
> could make default for an "ordered" dictionary. Suppose you say it should
> be ordered by insertion order, so
> d = OrderedDict(); d[1]='one'; d[2]='two' =>> list(d) => [1, 2]
> ok, now we do d[1]='ein' and what is the order? list(d) => [2, 1] ??
> Or do replacements not count as "insertions"?
"Insertion-order" is not a good term. For a dictionary {key:value} pair
creation, updating and deletion are possible modifying operations. I
don't see how deletion influences the order so creation and updating
remains. Therefore you have to select beween a creation_order and an
update_order. This can be reduced to one additional keyword e.g.
"creation_order" that is assigned a boolean value which is true by
default.
> The devil is always going
> to be in the details. Maybe you want a model that works more like a list
> of key:value pairs with just optimized access to a pair by key name as
> well as position in the list. Or maybe you want to permit append and
> NOT prevent [('a',1), ('a':2)] and maybe d['a'] => [1, 2] ???
As far as I understand the requirement an odict provides the same
interface as a dict. The only difference is a certain order of the keys
that is induced by operations on a dict and cannot be established by
properties of the keys ( or values ) itself.
> Note that is isn't hard to snap a few pieces together to make an ordered
> dict to your own specs. But IMO it belongs in pyPI or such, not in the system
> library. At least until it gets a lot of mileage -- and MMV ;-)
It's also not very hard to write a hex2ascii converter. That's the
reason why 20 incompatible versions of it ( coded in C ) exists in my
department ;)
Kay
PS. Here is some attempt of my own to implement an odict, following the
discussion here.
The implementation highlights just the model and is incomplete:
class odict(dict):
def __init__(self, create_order = True):
dict.__init__(self)
self.create_order = create_order
self.__cnt = 0
def __setitem__(self, key, value):
val = dict.get(self,key)
if val and self.create_order:
dict.__setitem__(self, key, (val[0], value))
else:
self.__cnt+=1
dict.__setitem__(self, key, (self.__cnt, value))
def __getitem__(self, key):
return dict.__getitem__(self, key)[1]
def values(self):
return list(zip(*sorted(dict.values(self)))[1])
def keys(self):
ks = [(dict.get(self,k)[0],k) for k in dict.keys(self)]
return list(zip(*sorted(ks))[1])
>>> od = odict()
>>> od["a"] = 0
>>> od["b"] = 8
>>> od.keys()
["a", "b"]
>>> od = odict(create_order = False)
>>> od["a"] = 1
>>> od["b"] = 2
>>> od["a"] = 3
>>> od.keys()
["b", "a"]
Christoph Zwerschke
which way to expose the keys and to mix in list methods for ordered
dictionaries.
In http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/107747
the keys are exposed via the keys() method which is bad. It should be a
copy only, like for ordinary dicts (one comment also mentions that).
In Foord/Larosa's odict, the keys are exposed as a public member which
also seems to be a bad idea ("If you alter the sequence list so that it
no longer reflects the contents of the dictionary, you have broken your
OrderedDict").
I think it would be probably the best to hide the keys list from the
public, but to provide list methods for reordering them (sorting,
slicing etc.).
For instance:
d1 = OrderedDict( (1, 11), (2, 12), 3, 13) )
d1[1:] ==> OrderedDict( (2, 12), 3, 13) )
d1[0] + d1[2] ==> OrderedDict( (1, 11), (3, 13) )
d1.reverse() ==> OrderedDict( (3, 13), (2, 12), 1, 11) )
d1.insert(1, (4, 14))
==> OrderedDict( (1, 11), (4, 14), (2, 12), 3, 13) )
etc.
But no other way to directly manipulate the keys should be provided.
-- Christoph
Christoph Zwerschke
>> I still believe that the concept of an "ordered dictionary" ("behave
>> like dict, only keep the order of the keys") is intuitive and doesn't
>> give you so much scope for ambiguity.
> Sure. Others think so too. The problem is that if you and these other
> people actually write down exactly how this unambigous ordered dict
> should behave, you'll end up with a dozen different sets of semantics,
> which can be divided in at least three main groups.
That's the point where I dare to disagree. As I have pointed out in
another posting in this thread, all other implementations have the same
semantics for the basic behavior. I cannot see three different groups.
Again, what's so surprising as the "natural" semantics described here:
http://mail.python.org/pipermail/python-dev/2005-March/052041.html
-- Christoph
Carsten Haese
> In Foord/Larosa's odict, the keys are exposed as a public member which
> also seems to be a bad idea ("If you alter the sequence list so that it
> no longer reflects the contents of the dictionary, you have broken your
> OrderedDict").
That could easily be fixed by making the sequence a "managed property"
whose setter raises a ValueError if you try to set it to something
that's not a permutation of what it was.
> d1[0] + d1[2] ==> OrderedDict( (1, 11), (3, 13) )
What do you think you're doing here?
-Carsten
Christoph Zwerschke
> I don't think it's intuitive if you can't describe it without
> contradicting yourself. If the order of the keys really were the order
> in which they were first seen by the dictionary, deleting and recreating
> a key should maintain its original position.
Admitted that description was not perfect (anyway it was not mine ;-)).
If a key is deleted, it is deleted. I don't think it is intuitive to
expect that the dict "remembers" a deleted item. If it is added again,
it is like it is seen for the first time and thus appended.
I don't think your argument viliates what I said in my last post.
-- Chris
Christoph Zwerschke
>>like dict, only keep the order of the keys") is intuitive and doesn't
>>give you so much scope for ambiguity. But probably I need to work on an
>>implementation to become more clear about possible hidden subtleties.
Bengt Richter schrieb:
> Does that mean that the Larosa/Foord odict.py implementation in PyPI
> does not do what you want?
Basically, it is what I had in mind. But I'm now seeing some things that
could be improved/supplemented, e.g.
- performance improvements
- the internal keys list should be hidden
- list methods should be mixed in instead
-- Christoph
Christoph Zwerschke
> That could easily be fixed by making the sequence a "managed property"
> whose setter raises a ValueError if you try to set it to something
> that's not a permutation of what it was.
Ok, a managed attribute would be an option. You would allow people to do
what they want with the sequence, and then fix the dictionary
accordingly (if items were deleted from the sequence, they are deleted
from the dictionary, it items were added which are not in the directory,
a ValueError is raised etc.).
But probably it is still better to hide the sequence and instead of
letting people do list operations like sort() on the odict.sequence, let
them do these operations on odict directly.
>>d1[0] + d1[2] ==> OrderedDict( (1, 11), (3, 13) )
> What do you think you're doing here?
Sorry, what I meant was
d1[0:0] + d1[2:2] ==> OrderedDict( (1, 11), (3, 13) )
Ordered dictionaries could allow slicing and concatenation.
-- Christoph
Christoph Zwerschke
Oops, sorry, that was nonsense again. I meant
d1[0:1] + d1[1:2] ==> OrderedDict( (1, 11), (3, 13) )
> Ordered dictionaries could allow slicing and concatenation.
Some operations such as concatenation need of course special
considerations, since the keys must stay unique. A concatenation of
ordered dicts with overlapping keys should probably give an IndexError.
-- Christoph
Bengt Richter
Those are zero-length slices in normal notation. ITYM [0:1] and [2:3]?
>
Note that you'd have to define addition as possible replacement,
if all the keys happened to match. Or pure concat if none matched,
and variations mixing both ways.
But with the current version you can already write that as
OrderedDict(d1.items()[0:1]+d2.items()[2:3])
you just want the sugar? d1+d2 would be like using [:] in the above line
Not a biggie to do.
Regards,
Bengt Richter
Bengt Richter
Since the OrderedDict constructor takes a sequence of tuples as a legitimate
argument, you can always create an ordered dict from an unordered by getting
unordered_dict.items() and sorting or ordering them any way you want and calling
the OrderedDict constructor. Ditto for ordered dicts, since they give your their
ordered items with the items() method as a start. I guess one could pass a
key=fun keyword arg to the OrderedDict constuctor to imply a pre-construction sort.
Regards,
Bengt Richter
Bengt Richter
>Bengt Richter wrote:
>> On Mon, 21 Nov 2005 01:27:22 +0100, Christoph Zwerschke <ci...@online.de> wrote:
>
>> Note that is isn't hard to snap a few pieces together to make an ordered
>> dict to your own specs. But IMO it belongs in pyPI or such, not in the system
>> library. At least until it gets a lot of mileage -- and MMV ;-)
>
>It's also not very hard to write a hex2ascii converter. That's the
>reason why 20 incompatible versions of it ( coded in C ) exists in my
>department ;)
Bicycle shed effect, I guess ;-)
>
>Kay
>
>PS. Here is some attempt of my own to implement an odict, following the
>discussion here.
>The implementation highlights just the model and is incomplete:
>
This is essentially the tack I took in modifying odict.py, except I
added optional caching of sorted items, and other minor differences.
>class odict(dict):
> def __init__(self, create_order = True):
> dict.__init__(self)
> self.create_order = create_order
> self.__cnt = 0
>
> def __setitem__(self, key, value):
> val = dict.get(self,key)
> if val and self.create_order:
> dict.__setitem__(self, key, (val[0], value))
> else:
> self.__cnt+=1
> dict.__setitem__(self, key, (self.__cnt, value))
>
> def __getitem__(self, key):
> return dict.__getitem__(self, key)[1]
>
> def values(self):
> return list(zip(*sorted(dict.values(self)))[1])
maybe more directly
return [v for i,v in sorted(dict.values(self))]
>
> def keys(self):
> ks = [(dict.get(self,k)[0],k) for k in dict.keys(self)]
> return list(zip(*sorted(ks))[1])
or (untested)
def keys(self):
return [k for k,v in sorted(dict.items(self), key=operator.itemgetter(1))]
def items(self):
return [(k,v[1]) for k,v in sorted(dict.items(self), key=operator.itemgetter(1))]
>
>>>> od = odict()
>>>> od["a"] = 0
>>>> od["b"] = 8
>>>> od.keys()
>["a", "b"]
>
>>>> od = odict(create_order = False)
>>>> od["a"] = 1
>>>> od["b"] = 2
>>>> od["a"] = 3
>>>> od.keys()
>["b", "a"]
>
Regards,
Bengt Richter
bon...@gmail.com
Bengt Richter wrote:
> On 22 Nov 2005 03:07:47 -0800, "bon...@gmail.com" <bon...@gmail.com> wrote:
>
> >
> >Bengt Richter wrote:
> >> Ok, so if not in the standard library, what is the problem? Can't find what
> >> you want with google and PyPI etc.? Or haven't really settled on what your
> >> _requirements_ are? That seems to be the primary problem people who complain
> >> with "why no sprollificator mode?" questions. They don't know what they really
> >> mean when it comes down to a DYFR (Define Your Felicitous Requirements) challenge.
> >> So DYFR ;-)
> >Beat me. I am not the one asking the question.
> Sorry, I thought you wanted an ordered dict too.
order but I don't want or need a standard one as I don't think there is
ONE implementation of it. My original post was a response to the
question "why do one want ordered dict", in the tone of "there is no
way one wants it".
>
> >
> >> >> > parsing or not parsing is not the point, and parsing/converting is
> >> >> > still "create a new view" of an existing data structure.
> >> >>
> >> So you'd like the mechanics to be automated and hidden? Then you need to
> >> DYFR for using the black box you want. Methods, semantics.
> >Lose you. don't know what you want to say.
> >
> I like solving problems. I just get frustrated when people don't focus on getting
> the problem defined, which IME is 2/3 of the way to a solution. I don't mind,
> in fact enjoy, rambling musings, but if someone seems actually to want a solution
> for something, I like to try to realize it concretely.
I tried to define the problem, and how I solve it(if it helps to convey
the message), but was told you don't have the problem in the first
place.
Tom Anderson
> On Tue, 2005-11-22 at 14:37, Christoph Zwerschke wrote:
>
>> In Foord/Larosa's odict, the keys are exposed as a public member which
>> also seems to be a bad idea ("If you alter the sequence list so that it
>> no longer reflects the contents of the dictionary, you have broken your
>> OrderedDict").
>
> That could easily be fixed by making the sequence a "managed property"
> whose setter raises a ValueError if you try to set it to something
> that's not a permutation of what it was.
I'm not a managed property expert (although there's a lovely studio in
Bayswater you might be interested in), but how does this stop you doing:
my_odict.sequence[0] = Shrubbery()
Which would break the odict good and hard.
tom
--
When I see a man on a bicycle I have hope for the human race. --
H. G. Wells
Bengt Richter
>>> from odictb import OrderedDict
>>> d1 = OrderedDict([(1, 11), (2, 12), (3, 13)])
>>> d1
{1: 11, 2: 12, 3: 13}
>>> d1[1:]
{2: 12, 3: 13}
>>> d1[0:1] + d1[2:3]
{1: 11, 3: 13}
>>> d1.reverse()
>>> d1
{3: 13, 2: 12, 1: 11}
>>> d1.insert(1, (4,14))
>>> d1
{3: 13, 4: 14, 2: 12, 1: 11}
>>> d1.items()
[(3, 13), (4, 14), (2, 12), (1, 11)]
>>> d1.keys()
[3, 4, 2, 1]
>>> d1.values()
[13, 14, 12, 11]
>>> d1[1:2]
{4: 14}
>>> d1[-1:]
{1: 11}
Que mas?
Regards,
Bengt Richter
Tom Anderson
> One implementation detail that I think needs further consideration is in
> which way to expose the keys and to mix in list methods for ordered
> dictionaries.
>
> In Foord/Larosa's odict, the keys are exposed as a public member which
> also seems to be a bad idea ("If you alter the sequence list so that it
> no longer reflects the contents of the dictionary, you have broken your
> OrderedDict").
>
> I think it would be probably the best to hide the keys list from the public,
> but to provide list methods for reordering them (sorting, slicing etc.).
I'm not too keen on this - there is conceptually a list here, even if it's
one with unusual constraints, so there should be a list i can manipulate
in code, and which should of course be bound by those constraints.
I think the way to do it is to have a sequence property (which could be a
managed attribute to prevent outright clobberation) which walks like a
list, quacks like a list, but is in fact a mission-specific list subtype
whose mutator methods zealously enforce the invariants guaranteeing the
odict's integrity.
I haven't actually tried to write such a beast, so i don't know if this is
either of possible and straightforward.