By cs95


2019-01-03 18:54:20 8 Comments

Are for loops really "bad"? If not, in what situation(s) would they be better than using a more conventional "vectorized" approach?1

I am familiar with the concept of "vectorization", and how pandas employs vectorized techniques to speed up computation. Vectorized functions broadcast operations over the entire series or DataFrame to achieve speedups much greater than conventionally iterating over the data.

However, I am quite surprised to see a lot of code (including from answers on Stack Overflow) offering solutions to problems that involve looping through data using for loops and list comprehensions. The documentation and API say that loops are "bad", and that one should "never" iterate over arrays, series, or DataFrames. So, how come I sometimes see users suggesting loop-based solutions?


1 - While it is true that the question sounds somewhat broad, the truth is that there are very specific situations when for loops are usually better than conventionally iterating over data. This post aims to capture this for posterity.

1 comments

@cs95 2019-01-03 18:54:20

TLDR; No, for loops are not blanket "bad", at least, not always. It is probably more accurate to say that some vectorized operations are slower than iterating, versus saying that iteration is faster than some vectorized operations. Knowing when and why is key to getting the most performance out of your code. In a nutshell, these are the situations where it is worth considering an alternative to vectorized pandas functions:

  1. When your data is small (...depending on what you're doing),
  2. When dealing with object/mixed dtypes
  3. When using the str/regex accessor functions

Let's examine these situations individually.


Iteration v/s Vectorization on Small Data

Pandas follows a "Convention Over Configuration" approach in its API design. This means that the same API has been fitted to cater to a broad range of data and use cases.

When a pandas function is called, the following things (among others) must internally be handled by the function, to ensure working

  1. Index/axis alignment
  2. Handling mixed datatypes
  3. Handling missing data

Almost every function will have to deal with these to varying extents, and this presents an overhead. The overhead is less for numeric functions (for example, Series.add), while it is more pronounced for string functions (for example, Series.str.replace).

for loops, on the other hand, are faster then you think. What's even better is list comprehensions (which create lists through for loops) are even faster as they are optimized iterative mechanisms for list creation.

List comprehensions follow the pattern

[f(x) for x in seq]

Where seq is a pandas series or DataFrame column. Or, when operating over multiple columns,

[f(x, y) for x, y in zip(seq1, seq2)]

Where seq1 and seq2 are columns.

Numeric Comparison
Consider a simple boolean indexing operation. The list comprehension method has been timed against Series.ne (!=) and query. Here are the functions:

# Boolean indexing with Numeric value comparison.
df[df.A != df.B]                            # vectorized !=
df.query('A != B')                          # query (numexpr)
df[[x != y for x, y in zip(df.A, df.B)]]    # list comp

For simplicity, I have used the perfplot package to run all the timeit tests in this post. The timings for the operations above are below:

enter image description here

The list comprehension outperforms query for moderately sized N, and even outperforms the vectorized not equals comparison for tiny N. Unfortunately, the list comprehension scales linearly, so it does not offer much performance gain for larger N.

Note
It is worth mentioning that much of the benefit of list comprehension come from not having to worry about the index alignment, but this means that if your code is dependent on indexing alignment, this will break. In some cases, vectorised operations over the underlying NumPy arrays can be considered as bringing in the "best of both worlds", allowing for vectorisation without all the unneeded overhead of the pandas functions. This means that you can rewrite the operation above as

df[df.A.values != df.B.values]

Which outperforms both the pandas and list comprehension equivalents:

NumPy vectorization is out of the scope of this post, but it is definitely worth considering, if performance matters.

Value Counts
Taking another example - this time, with another vanilla python construct that is faster than a for loop - collections.Counter. A common requirement is to compute the value counts and return the result as a dictionary. This is done with value_counts, np.unique, and Counter:

# Value Counts comparison.
ser.value_counts(sort=False).to_dict()           # value_counts
dict(zip(*np.unique(ser, return_counts=True)))   # np.unique
Counter(ser)                                     # Counter

enter image description here

The results are more pronounced, Counter wins out over both vectorized methods for a larger range of small N (~3500).

Note
More trivia (courtesy @user2357112). The Counter is implemented with a C accelerator, so while it still has to work with python objects instead of the underlying C datatypes, it is still faster than a for loop. Python power!

Of course, the take away from here is that the performance depends on your data and use case. The point of these examples is to convince you not to rule out these solutions as legitimate options. If these still don't give you the performance you need, there is always cython and numba. Let's add this test into the mix.

from numba import njit, prange

@njit(parallel=True)
def get_mask(x, y):
    result = [False] * len(x)
    for i in prange(len(x)):
        result[i] = x[i] != y[i]

    return np.array(result)

df[get_mask(df.A.values, df.B.values)] # numba

enter image description here

Numba offers JIT compilation of loopy python code to very powerful vectorized code. Understanding how to make numba work involves a learning curve.


Operations with Mixed/object dtypes

String-based Comparison
Revisiting the filtering example from the first section, what if the columns being compared are strings? Consider the same 3 functions above, but with the input DataFrame cast to string.

# Boolean indexing with string value comparison.
df[df.A != df.B]                            # vectorized !=
df.query('A != B')                          # query (numexpr)
df[[x != y for x, y in zip(df.A, df.B)]]    # list comp

enter image description here

So, what changed? The thing to note here is that string operations are inherently difficult to vectorize. Pandas treats strings as objects, and all operations on objects fall back to a slow, loopy implementation.

Now, because this loopy implementation is surrounded by all the overhead mentioned above, there is a constant magnitude difference between these solutions, even though they scale the same.

When it comes to operations on mutable/complex objects, there is no comparison. List comprehension outperforms all operations involving dicts and lists.

Accessing Dictionary Value(s) by Key
Here are timings for two operations that extract a value from a column of dictionaries: map and the list comprehension. The setup is in the Appendix, under the heading "Code Snippets".

# Dictionary value extraction.
ser.map(operator.itemgetter('value'))     # map
pd.Series([x.get('value') for x in ser])  # list comprehension

enter image description here

Positional List Indexing
Timings for 3 operations that extract the 0th element from a list of columns (handling exceptions), map, str.get accessor method, and the list comprehension:

# List positional indexing. 
def get_0th(lst):
    try:
        return lst[0]
    # Handle empty lists and NaNs gracefully.
    except (IndexError, TypeError):
        return np.nan

ser.map(get_0th)                                          # map
ser.str[0]                                                # str accessor
pd.Series([x[0] if len(x) > 0 else np.nan for x in ser])  # list comp
pd.Series([get_0th(x) for x in ser])                      # list comp safe

Note
If the index matters, you would want to do:

pd.Series([...], index=ser.index)

When reconstructing the series.

enter image description here

List Flattening
A final example is flattening lists. This is another common problem, and demonstrates just how powerful pure python is here.

# Nested list flattening.
pd.DataFrame(ser.tolist()).stack().reset_index(drop=True)  # stack
pd.Series(list(chain.from_iterable(ser.tolist())))         # itertools.chain
pd.Series([y for x in ser for y in x])                     # nested list comp

enter image description here

Both itertools.chain.from_iterable and the nested list comprehension are pure python constructs, and scale much better than the stack solution.

These timings are a strong indication of the fact that pandas is not equipped to work with mixed dtypes, and that you should probably refrain from using it to do so. Wherever possible, data should be present as scalar values (ints/floats/strings) in separate columns.

Lastly, the applicability of these solutions depend widely on your data. So, the best thing to do would be to test these operations on your data before deciding what to go with. Notice how I have not timed apply on these solutions, because it would skew the graph (yes, it's that slow).


Regex Operations, and .str Accessor Methods

Pandas can apply regex operations such as str.contains, str.extract, and str.extractall, as well as other "vectorized" string operations (such as str.split, str.find,str.translate`, and so on) on string columns. These functions are slower than list comprehensions, and are meant to be more convenience functions than anything else.

It is usually much faster to pre-compile a regex pattern and iterate over your data with re.compile (also see Is it worth using Python's re.compile?). The list comp equivalent to str.contains looks something like this:

p = re.compile(...)
ser2 = pd.Series([x for x in ser if p.search(x)])

Or,

ser2 = ser[[bool(p.search(x)) for x in ser]]

If you need to handle NaNs, you can do something like

ser[[bool(p.search(x)) if pd.notnull(x) else False for x in ser]]

The list comp equivalent to str.extract (without groups) will look something like:

df['col2'] = [p.search(x).group(0) for x in df['col']]

If you need to handle no-matches and NaNs, you can use a custom function (still faster!):

def matcher(x):
    m = p.search(str(x))
    if m:
        return m.group(0)
    return np.nan

df['col2'] = [matcher(x) for x in df['col']]

The matcher function is very extensible. It can be fitted to return a list for each capture group, as needed. Just extract query the group or groups attribute of the matcher object.

For str.extractall, change p.search to p.findall.

String Extraction
Consider a simple filtering operation. The idea is to extract 4 digits if it is preceded by an upper case letter.

# Extracting strings.
p = re.compile(r'(?<=[A-Z])(\d{4})')
def matcher(x):
    m = p.search(x)
    if m:
        return m.group(0)
    return np.nan

ser.str.extract(r'(?<=[A-Z])(\d{4})', expand=False)   #  str.extract
pd.Series([matcher(x) for x in ser])                  #  list comprehension

enter image description here

More Examples
Full disclosure - I am the author (in part or whole) of these posts listed below.


Conclusion

As shown from the examples above, iteration shines when working with small rows of DataFrames, mixed datatypes, and regular expressions.

The speedup you get depends on your data and your problem, so your mileage may vary. The best thing to do is to carefully run tests and see if the payout is worth the effort.

The "vectorized" functions shine in their simplicity and readability, so if performance is not critical, you should definitely prefer those.

Another side note, certain string operations deal with constraints that favour the use of NumPy. Here are two examples where careful NumPy vectorization outperforms python:

Additionally, sometimes just operating on the underlying arrays via .values as opposed to on the Series or DataFrames can offer a healthy enough speedup for most usual scenarios (see the Note in the Numeric Comparison section above). So, for example df[df.A.values != df.B.values] would show instant performance boosts over df[df.A != df.B]. Using .values may not be appropriate in every situation, but it is a useful hack to know.

As mentioned above, it's up to you to decide whether these solutions are worth the trouble of implementing.


Appendix: Code Snippets

import perfplot  
import operator 
import pandas as pd
import numpy as np
import re

from collections import Counter
from itertools import chain

# Boolean indexing with Numeric value comparison.
perfplot.show(
    setup=lambda n: pd.DataFrame(np.random.choice(1000, (n, 2)), columns=['A','B']),
    kernels=[
        lambda df: df[df.A != df.B],
        lambda df: df.query('A != B'),
        lambda df: df[[x != y for x, y in zip(df.A, df.B)]],
        lambda df: df[get_mask(df.A.values, df.B.values)]
    ],
    labels=['vectorized !=', 'query (numexpr)', 'list comp', 'numba'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N'
)

# Value Counts comparison.
perfplot.show(
    setup=lambda n: pd.Series(np.random.choice(1000, n)),
    kernels=[
        lambda ser: ser.value_counts(sort=False).to_dict(),
        lambda ser: dict(zip(*np.unique(ser, return_counts=True))),
        lambda ser: Counter(ser),
    ],
    labels=['value_counts', 'np.unique', 'Counter'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N',
    equality_check=lambda x, y: dict(x) == dict(y)
)

# Boolean indexing with string value comparison.
perfplot.show(
    setup=lambda n: pd.DataFrame(np.random.choice(1000, (n, 2)), columns=['A','B'], dtype=str),
    kernels=[
        lambda df: df[df.A != df.B],
        lambda df: df.query('A != B'),
        lambda df: df[[x != y for x, y in zip(df.A, df.B)]],
    ],
    labels=['vectorized !=', 'query (numexpr)', 'list comp'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N',
    equality_check=None
)

# Dictionary value extraction.
ser1 = pd.Series([{'key': 'abc', 'value': 123}, {'key': 'xyz', 'value': 456}])
perfplot.show(
    setup=lambda n: pd.concat([ser1] * n, ignore_index=True),
    kernels=[
        lambda ser: ser.map(operator.itemgetter('value')),
        lambda ser: pd.Series([x.get('value') for x in ser]),
    ],
    labels=['map', 'list comprehension'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N',
    equality_check=None
)

# List positional indexing. 
ser2 = pd.Series([['a', 'b', 'c'], [1, 2], []])        
perfplot.show(
    setup=lambda n: pd.concat([ser2] * n, ignore_index=True),
    kernels=[
        lambda ser: ser.map(get_0th),
        lambda ser: ser.str[0],
        lambda ser: pd.Series([x[0] if len(x) > 0 else np.nan for x in ser]),
        lambda ser: pd.Series([get_0th(x) for x in ser]),
    ],
    labels=['map', 'str accessor', 'list comprehension', 'list comp safe'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N',
    equality_check=None
)

# Nested list flattening.
ser3 = pd.Series([['a', 'b', 'c'], ['d', 'e'], ['f', 'g']])
perfplot.show(
    setup=lambda n: pd.concat([ser2] * n, ignore_index=True),
    kernels=[
        lambda ser: pd.DataFrame(ser.tolist()).stack().reset_index(drop=True),
        lambda ser: pd.Series(list(chain.from_iterable(ser.tolist()))),
        lambda ser: pd.Series([y for x in ser for y in x]),
    ],
    labels=['stack', 'itertools.chain', 'nested list comp'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N',    
    equality_check=None

)

# Extracting strings.
ser4 = pd.Series(['foo xyz', 'test A1234', 'D3345 xtz'])
perfplot.show(
    setup=lambda n: pd.concat([ser4] * n, ignore_index=True),
    kernels=[
        lambda ser: ser.str.extract(r'(?<=[A-Z])(\d{4})', expand=False),
        lambda ser: pd.Series([matcher(x) for x in ser])
    ],
    labels=['str.extract', 'list comprehension'],
    n_range=[2**k for k in range(0, 15)],
    xlabel='N',
    equality_check=None
)

@GZ0 2019-06-09 04:49:52

pd.Series and pd.DataFrame now support construction from iterables. That means one can simply pass a Python generator to the constructor functions rather than needing to construct a list first (using list comprehensions), which could be slower in many cases. However, the size of the generator output cannot be determined beforehand. I'm not sure how much time / memory overhead that would cause.

@cs95 2019-06-09 04:51:52

@GZ0 IIRC, accepting iterators is a more recent addition to the API. As for "that means one can simply pass a Python generator to the constructor functions rather than needing to construct a list first", I disagree. Memory maybe, performance no. In my experience it usually almost always faster to generate the list and pass it on. List comps FTW.

@GZ0 2019-06-09 05:00:09

@cs95 Why don't you think generators can lead to performance gain? Or had you done any tests on that?

@cs95 2019-06-09 05:11:43

@GZ0 I didn't say generators do not lead to a performance gain, I'm saying you don't gain as much as you would using list comprehensions. There are overheads associated with initialising genexps, maintaining state, etc, which list comps don't have. There are a lot of examples here with list comprehensions that you can time against generator comprehensions to compare. Here's an example you can run right now: ser = pd.Series([['a', 'b', 'c'], ['d', 'e'], ['f', 'g']] * 10000) ; %timeit pd.Series(y for x in ser for y in x) ; %timeit pd.Series([y for x in ser for y in x])

@GZ0 2019-06-09 05:35:59

@cs95 You are probably right. I just had a quick look into the constructor code of pd.Series and it turns out that if the size of the iterable output cannot be determined a list would be constructed from the iterable anyway. So in most cases there would be no gains to use generators. However when I ran the examples you gave the generator version is actually slower. I'm not sure where the difference comes from.

@GZ0 2019-06-09 05:52:41

@cs95 After some investigation I think the difference is primarily attributed to list() calls being slower than list comprehensions.

@cs95 2019-06-09 05:55:34

@GZ0 That, plus the overhead of bookkeeping, maintaining state, etc. There is no gain whatsoever for any function that has to construct the list before using its values, as Series and DataFrame will do. str.join is a case in point.

@GZ0 2019-06-09 06:15:33

@cs95 str.join is similar to pd.Series in that it also needs to know the total size of input before constructing the output. So if pandas and numpy use primarily fixed-sized arrays for the underlying implementation, which I assume is the case, then they have to know the total size beforehand.

@GZ0 2019-06-09 07:21:06

@cs95 Note that for known-size iterables, it would be faster to construct pd.Series from them directly rather than converting them to lists, e.g. pd.Series(range(10000)), pd.Series("a" * 10000), and pd.Series(pd.Index(range(10000))) would be much faster than their list counterparts (the last one was even slightly faster than pd.Index.to_series.

Related Questions

Sponsored Content

5 Answered Questions

[SOLVED] Creating an empty Pandas DataFrame, then filling it?

1 Answered Questions

Speed up applying function to a list of pandas dataframes

1 Answered Questions

[SOLVED] How to groupby, loop and get result as a dataframe in panda

  • 2019-02-18 15:49:04
  • latmos
  • 82 View
  • 0 Score
  • 1 Answer
  • Tags:   python pandas

1 Answered Questions

[SOLVED] Vectorized look-up of values in Pandas dataframe

2 Answered Questions

[SOLVED] Scalar-valued isnull()/isnan()/isinf()

1 Answered Questions

[SOLVED] python pandas loop append dataframe

1 Answered Questions

[SOLVED] Remove for loop in python pandas find and replace text

1 Answered Questions

[SOLVED] Faster For Loop to Manipulate Data in Pandas

  • 2016-09-12 13:58:27
  • NickBraunagel
  • 283 View
  • 1 Score
  • 1 Answer
  • Tags:   python pandas numpy

Sponsored Content