retain unique combinations from a list of tuples in python

Question

I have a list of tuples that contain int and str values in them:

Listex:
[23,'item1','item2']
[23,'item2','item1']
[67,'item3','item2']
[55,'item3','item4']
[67,'item2','item3']
[55,'item4','item3']

What I need is to get unique tuples:

Listex output:
[23,'item1','item2']
[67,'item2','item3']
[55,'item3','item4']

I have tried sort() but I get the error that '<' cannot be used with str and int. Appreciate any help in this regard. Thank you in advance.

Answer 1

First of all usefrozenset so that comparisons wont take into account the order of the strings, then use a set to remove duplicates, finally convert back to list of lists:

In [1]: data = [[23,'item1','item2'], 
   ...: [23,'item2','item1'], 
   ...: [67,'item3','item2'], 
   ...: [55,'item3','item4'], 
   ...: [67,'item2','item3'], 
   ...: [55,'item4','item3']]                                                                                                                                                                                                                                                            

In [2]: list(map(list, set(map(frozenset, data))))                                                                                                                                                                                                                                       
Out[2]: [[67, 'item2', 'item3'], ['item1', 'item2', 23], ['item3', 'item4', 55]]

You cannot directly apply set to your original collection because set requires its elements to be hashable , which list s are not. Converting the inner lists into frozenset makes them hashable and has the added bonus of being order-independent. Using a tuple would make the elements hashable but order of the elements would matter.

Depending on what you want to do with the data maybe you can simply use a set of frozensets and avoid the lists altogether:

In [3]: set(map(frozenset, data))                                                                                                                                                                                                                                                        
Out[3]: 
{frozenset({67, 'item2', 'item3'}),
 frozenset({23, 'item1', 'item2'}),
 frozenset({55, 'item3', 'item4'})}

---

If you want to keep the number as the first element you can replace list with some function like:

In [5]: def keep_number_first(seq): 
   ...:     return sorted(seq, key=lambda x: 1 if not isinstance(x, (int, float)) else 0) 
   ...:           

which sorts by type:

In [6]: list(map(keep_number_first, set(map(frozenset, data))))                                                                                                                                                                                                                          
Out[6]: [[67, 'item2', 'item3'], [23, 'item1', 'item2'], [55, 'item3', 'item4']]

---

This said: if you have to keep data structured in this way, IMHO, it's probably better to come up with your own class or use a namedtuple and give some meaning to the different pieces. For example instead of having [23, 'item1', 'item2'] why not (23, ['item1', 'item2']) ?