在有序字典中将键元素添加到键值

Question

I have an ordered dictionary:

{'20140106': '82.1000000',
 '20140217': '77.0300000',
 '20140224': '69.6200000',
 '20140310': '46.3300000',
 '20140414': '49.3800000',

Keys are dates: yyyy-mm-dd and i want to store all values of one month in one key:

{'20140106': '82.1000000',
 '20140217': ['77.0300000','69.6200000',]
 '20140310': '46.3300000',
 '20140414': '49.3800000',

I realized this not very beutiful solution:

val = []   #list, where i put all values for an exact month
previus_key = list(collection.keys())[0]
val.append(collection.get(previus_key))

for key in list(collection.keys())[1:]:
    if (key[4:6]==previus_key[4:6]):    #if it's the same month
        val.append(list(collection.pop(key)))             #remember the value and delete an item
    else:
        collection.update({previus_key:val})   #that means, we jumped to another month, so lets update the values for previus month 
        previus_key = key            #now, start the same algorihtm for a new month.
    val = collection.get(previus_key)

But, i get an error: 'str' object has no attribute 'append' for line val.append(list(collection.pop(key))) . I researched it, and came to conclusion, that it must not be here! Because, i append a list to a list! So, please, point to my mistake and give advice, how could i make the code more beutiful. Thanks!

Answer 1

I think the issue lies in the last line of your code. You do assign a string to the variable var there.

EDIT : Here is a suggestion which stays close to your original code.

new_collection = {}
for key in collection:
    month = key[:6]
    if new_collection.get(month) != None:
        new_collection[month].append(collection[key])
    else:
        new_collection[month] = [collection[key]]

Three things : 1) Here the result is a new dictionary instead of the same instance, as I find it generally preferable. 2) The keys are only the year and the month, as the day is irrelevant. 3) All the values are lists, instead of a mix of strings and lists.

Answer 2

Here is a better approach. Given a dict of type {str:str}, we can turn the value into a list on inserts while retaining all the functionality of dict by extending the builtin dict class.

Here is a demonstration.

>>> class CustomDict(dict):
...     def insert(self, key, val): # add a new method, `insert`
...             if self.get(key):
...                     if not isinstance(self.get(key), list):
...                             lst = []
...                             lst.append(self.get(key))
...                             lst.append(val)
...                             self[key] = lst[:] # copy list
...                     else:
...                             self[key].append(val)
...             else:
...                     self[key] = val
...
>>> vars = {'19900326': '372.99101',
...         '19730529': '291.38291',
...         '19430122': '81.248291',
...         '19930227': '192.32919',}
>>> vars2 = CustomDict(vars)
>>> vars2
{'19730529': '291.38291',
 '19900326': '372.99101',
 '19430122': '81.248291',
 '19930227': '192.32919'}
>>> vars2.insert('19730529', '102.88391')
>>> vars2
{'19730529': ['291.38291', '102.88391'],
 '19900326': '372.99101',
 '19430122': '81.248291',
 '19930227': '192.32919'}

In your case, you'll want to inherit from collections.ordereddict

As far as the logic of wanting to store all the pieces of data from the same month in the same list, you can pretty easily code that in there. Just to be sure though, it looks like you're currently disregarding year entirely. Are you sure all you care about is the same month, even if the data comes from different years? If not, you should change your logic so that you can differentiate which year.

Somebody suggested a better representation of the date, as a 3-tuple (year, month, day) . This is immutable, so it can be your dict key. (Lists are mutable and cannot be dict keys).

You could add the month-grouping logic to the insert method, or you could move that logic elsewhere in your code.

Answer 3

I don't see any advantage of using strings here as keys to represent the date. This is my solution, but I hope it won't break your current code.

Well, your code shows you're doing so much nesting there:

 val.append(list(collection.pop(key)))

And:

Keys are dates: yyyy-mm-dd and i want to store all values of one month in one key:

Here's a way to do it: Use tuples for this purpose:

mydict = {(year, month, day): value}

this way not only you're simplifying the problem, but you're also making your code more readable:

for date in mydict: 
    year  = date[0] 
    month = date[1] 
         ...
    ...process data...

Though keys can only be immutable objects, hence tuples are immutable objects.

Answer 4

Why not make use of reduce and filter function. reduce will apply a function to every item in the iterated object, and you can also specify the initial value for reduce

from collections import OrderedDict

a = {'20140106': '82.1000000', '20140217': '77.0300000', '20140224': '69.6200000', '20140310': '46.3300000', '20140414': '49.3800000'}


def aggr(result, item):
    key = filter(lambda x: x[:6] == item[0][:6], result.keys())
    if not key:
        result[item[0]] = [item[1]]
    else:
        result[key[0]].append(item[1])
    return result


r = OrderedDict()
print reduce(aggr, a.items(), r)