Simplified multi-column sorting of lists of tuples, dicts, lists or objects that are NoneType safe.
python3 -m pip install multisort
None
Average over 10 iterations with 1000 rows.
Test | Secs |
---|---|
superfast | 0.0005 |
multisort | 0.0035 |
pandas | 0.0079 |
cmp_func | 0.0138 |
reversor | 0.037 |
Hands down the fastest is the superfast
methdology shown below. You do not need this library to accomplish this as its just core python.
multisort
from this library gives reasonable performance for large data sets; eg. its better than pandas up to about 5,500 records. It is also much simpler to read and write, and it has error handling that does its best to give useful error messages.
If your data may contain None, it would be wise to ensure your sort algorithm is tuned to handle them. This is because sorted uses <
comparisons; which is not supported by NoneType
. For example, the following error will result: TypeError: '>' not supported between instances of 'NoneType' and 'str'
. All examples given on this page are tuned to handle None
values.
Method | Descr | Notes |
---|---|---|
multisort | Simple one-liner designed after multisort example from python docs |
Second fastest of the bunch but most configurable and easy to read. |
cmp_func | Multi column sorting in the model java.util.Comparator |
Reasonable speed |
superfast | NoneType safe sample implementation of multi column sorting as mentioned in example from python docs | Fastest by orders of magnitude but a bit more complex to write. |
For data:
rows_before = [
{'idx': 0, 'name': 'joh', 'grade': 'C', 'attend': 100}
,{'idx': 1, 'name': 'jan', 'grade': 'a', 'attend': 80}
,{'idx': 2, 'name': 'dav', 'grade': 'B', 'attend': 85}
,{'idx': 3, 'name': 'bob' , 'grade': 'C', 'attend': 85}
,{'idx': 4, 'name': 'jim' , 'grade': 'F', 'attend': 55}
,{'idx': 5, 'name': 'joe' , 'grade': None, 'attend': 55}
]
Sort rows_before by grade, descending, then attend, ascending and put None first in results:
from multisort import multisort, mscol
rows_sorted = multisort(rows_before, [
mscol('grade', reverse=False),
'attend'
])
-or- without mscol
from multisort import multisort
rows_sorted = multisort(rows_before, [
('grade', False),
'attend'
])
Sort rows_before by grade, descending, then attend and call upper() for grade:
from multisort import multisort, mscol
rows_sorted = multisort(rows_before, [
mscol('grade', reverse=False, clean=lambda s: None if s is None else s.upper()),
'attend'
])
-or- without mscol
from multisort import multisort
rows_sorted = multisort(rows_before, [
('grade', False, lambda s: None if s is None else s.upper()),
'attend'
])
multisort
parameters:
option | dtype | description |
---|---|---|
rows |
int or str | Key to access data. int for tuple or list |
spec |
str, int, list | Sort specification. Can be as simple as a column key / index or mscol |
reverse |
bool | Reverse order of final sort (defalt = False) |
spec
entry options:
option | position | dtype | description |
---|---|---|---|
key |
0 | int or str | Key to access data. int for tuple or list |
reverse |
1 | bool | Reverse sort of column |
clean |
2 | func | Function / lambda to clean the value. These calls can cause a significant slowdown. |
default |
3 | any | Value to substitute if required==False and key does not exist or None is found. Can be used to achive similar functionality to pandas na_position |
required |
4 | bool | Default True. If False, will substitute None or default if key not found (not applicable for list or tuple rows) |
* spec
entries can be passed as:
type | description |
---|---|
String |
Column name |
tuple |
Tuple of 1 or more spec options in their order as listed (see position ) |
mscol() |
Importable helper to aid in readability. Suggested for three or more of the options. |
Sort rows_before by grade, descending, then attend and call upper() for grade:
def cmp_student(a,b):
k='grade'; va=a[k]; vb=b[k]
if va != vb:
if va is None: return -1
if vb is None: return 1
return -1 if va > vb else 1
k='attend'; va=a[k]; vb=b[k];
if va != vb: return -1 if va < vb else 1
return 0
rows_sorted = sorted(rows_before, key=cmp_func(cmp_student), reverse=True)
def key_grade(student):
grade = student['grade']
return grade is None, grade
def key_attend(student):
attend = student['attend']
return attend is None, attend
students_sorted = sorted(students, key=key_attend)
students_sorted.sort(key=key_grade, reverse=True)
For data:
class Student():
def __init__(self, idx, name, grade, attend):
self.idx = idx
self.name = name
self.grade = grade
self.attend = attend
def __str__(self): return f"name: {self.name}, grade: {self.grade}, attend: {self.attend}"
def __repr__(self): return self.__str__()
rows_before = [
Student(0, 'joh', 'C', 100)
,Student(1, 'jan', 'a', 80)
,Student(2, 'dav', 'B', 85)
,Student(3, 'bob', 'C', 85)
,Student(4, 'jim', 'F', 55)
,Student(5, 'joe', None, 55)
]
(Same syntax as with Dictionary example above)
Sort rows_before by grade, descending, then attend and call upper() for grade:
def cmp_student(a,b):
if a.grade != b.grade:
if a.grade is None: return -1
if b.grade is None: return 1
return -1 if a.grade > b.grade else 1
if a.attend != b.attend:
return -1 if a.attend < b.attend else 1
return 0
rows_sorted = sorted(rows_before, key=cmp_func(cmp_student), reverse=True)
For data:
rows_before = [
(0, 'joh', 'a' , 100)
,(1, 'joe', 'B' , 80)
,(2, 'dav', 'A' , 85)
,(3, 'bob', 'C' , 85)
,(4, 'jim', None , 55)
,(5, 'jan', 'B' , 70)
]
(COL_IDX, COL_NAME, COL_GRADE, COL_ATTEND) = range(0,4)
(Same syntax as with Dictionary example above)
Sort rows_before by grade, descending, then attend and call upper() for grade:
def cmp_student(a,b):
k=COL_GRADE; va=a[k]; vb=b[k]
if va != vb:
if va is None: return -1
if vb is None: return 1
return -1 if va > vb else 1
k=COL_ATTEND; va=a[k]; vb=b[k];
if va != vb:
return -1 if va < vb else 1
return 0
rows_sorted = sorted(rows_before, key=cmp_func(cmp_student), reverse=True)
multisort
can be used as a basic non-destructive sorter of lists where a traditional sort does so destructively:
_orig = [1, 4, 3, 6, 5]
_orig.sort(reverse=True)
This will sort _orig
in-place
In builtin python, to do a non-destructive sort it takes two lines:
_orig = [1, 4, 3, 6, 5]
_clone = [:]
_clone.sort(reverse=True)
With Multisort its just one line:
_orig = [1, 4, 3, 6, 5]
_sorted = multisort(_orig, reverse=True)
Where _orig
is left unchanged
Name | Descr | Other |
---|---|---|
tests/test_multisort.py | multisort unit tests | - |
tests/performance_tests.py | Tunable performance tests using asyncio | requires pandas |
tests/hand_test.py | Hand testing | - |