Welcome to Streams’s documentation!

Streams is an easy to use library to allow you to interpret your information as a data flow and process it in this way. It allows you parallel processing of a data flow and you can control it.

Actually Streams is dramatically inspired by Java 8 Stream API. Of course it is not a new beast in the zoo, I used the same approach in several projects before but this pattern goes to mainstream now and it is good to have it in Python too.

Just several examples to help you to feel what is it:

from requests import get
from operator import itemgetter

average_price = Stream(urls)                             \ # make a stream from the list of urls
    .map(requests.get, parallel=4)                       \ # do url  fetching in parallel. 4 threads / greenlets
    .map(lambda response: response.json()["model"])     \ # extract required field from JSON.
    .exclude(lambda model: model["deleted_at"] is None) \ # we need only active accounts so filter out deleted ones
    .map(itemgetter("price"))                            \ # get a price from the model
    .decimals()                                          \ # convert prices into decimals
    .average()                                           \ # calulate average from the list

And now let’s check the piece of code which does almost the same.

from concurrent.futures import ThreadPoolExecutor
from requests import get

with ThreadPoolExecutor(4) as pool:
    average_price = Decimal("0.00")
    fetched_items = pool.map(requests.get, urls)
    for response in fetched_items:
        model = response.json()["model]
        if model["deleted_at"] is None: continue
        sum_of += Decimal(model["price"])
    average_price /= len(urls)

So this is Stream approach. Streams is a lazy library and won’t do anything that is not needed. Let’s say you have urls as an iterator and it contains several billions of URLs that you can’t fit into the memory (ThreadPoolExecutor creates a list in the memory) or you want to build a pipeline of your data management and manipulate it according to some conditions, checkout Streams, maybe it will help you to create more accurate and maintainable code.

Just suppose Streams as a pipes from your *nix environment but migrated into Python. It also has some cool features you need to know about:

  • Laziness,
  • Small memory footprint even for massive data sets,
  • Automatic and configurable parallelization,
  • Smart concurrent pool management.

Terms of content

Installation

Install with Pip or easy_install.

$ pip install pystreams

or

$ easy_install pystreams

If you want, you can always download the latest bleeding edge from GitHub:

$ git clone git://github.com/9seconds/streams.git
$ cd streams
$ ./setup.py install

Streams supports Python 2.6, 2.7, 3.2, 3.3, 3.4 and PyPy. Probably other implementations like Jython or IronPython will work, but I haven’t tested them there.

User Guide

I supposed you’ve worked with Django and you’ve been using its ORM a lot. I will try to lead you to the idea of functional streams by example. Actually I did no Django for a while and syntax might be outdated a bit or I may confuse you so you are free to correct me through issue or pull request. Please do it, I appreciate your feedback.

If you didn’t work with any ORM just try to follow the idea, I will try to explain what is going on and things that really matter.

What is Stream?

Let’s go back to default Django example: libraries and books. Let’s assume that we have app up and running and it does some data management from your beloved database. Let’s say you want to fetch some recent books.

from library.models import Book

books = Book.objects.filter(pub_date__year=2014)

Good, isn’t it? You have a collection of models called Book which possibly presents books in your app. And you want to have only those which were published in 2014. Good, figured out. Let’s go further. Let’s say you want to be more specific and you want to have only bestsellers. It is ok.

from library.models import Book

books = Book.objects.filter(pub_date__year=2014)
bestsellers = books.order_by("-sales_count")[:10]

You can do it like this. But why is it better than this approach?

from operator import attrgetter
from library.models import Book

books = Book.objects.all()
books = [book for book in books if book.pub_date.year == 2014]
bestsellers = sorted(books, key=attrgetter("sales_count"), reverse=True)
bestsellers = bestsellers[:10]

You will get the same result, right? Actually no. Look, on filtering step you fetch all objects from the database and process them all. It is ok if you have a dozen of models in your database but it can be big bottleneck if your data is growing. That’s why everyone is trying to move as much filtering as possible into the database. Database knows how to manage your data accurately and what do to in the most efficient way. It will use indexes etc to speedup whole process and you do not need to do full scan everytime. It is best practice to fetch only that data you actually need from the database.

So instead of

SELECT * FROM book

you do

SELECT * FROM book
    WHERE EXTRACT(year FROM "pub_date") == 2014
    ORDER BY sales_count DESC
    LIMIT 10

Sounds legit. But let’s checkout how it looks like when do you work with ORM. Let’s go back to our example:

books = Book.objects.filter(pub_date__year=2014)
bestsellers = books.order_by("-sales_count")[:10]

or in a short way

bestsellers = Book.objects \
    .filter(pub_date__year=2014) \
    .order_by("-sales_count")[:10]

You may assume it like a data stream you are processing on every step. First you set initial source of data, this is Book.objects.all(). Good. You may consider it as an iterable flow of data and you apply processing functions on that stream, first if filtering, second is sorting, third is slicing. You process the flow, not every objects, this is crucial concept. Everytime after execution of some flow (or QuerySet) method you get another instance of the same flow but with your modifications.

You may suppose that Streams library to provide you the same functionality but for any iterable. Of course this is not that efficient as Django ORM which knows the context of database and helps you to execute your queries in the most efficient way.

How to use Streams

Now you got an idea of Streams: to manage data flow itself, not every component. You can build your own toy map/reduce stuff with it if you really need to have it. Our you can just filter and process your data to exclude some Nones etc in parallel or to have some generic way to do it. It is up to you, I’ll just show you some examples and if you want to have more information just go to the API documentation

So, for simplicity let’s assume that you have giant gzipped CSV, in 10 GB. And you can use only 1GB of your memory so it is not possible to put everything in memory at once. This CSV has 5 columns, author_id, book_name.

Yeah, books again. Why not?

So your boss asked you to implement function which will read this csvfile and do some optional filtering on it. Also you must fetch the data from predefined external sources, search on prices in different shops (Amazon at least) and write some big XML file with an average price.

I some explanation on the go.

from csv import reader
from gzip import open as gzopen
from collections import namedtuple
try:
    from xml.etree import cElementTree as ET
except ImportError:
    from xml.etree import ElementTree as ET
from streams import Stream
from other_module import shop_prices_fetch, author_fetch, publisher_fetch


def extract_averages(csv_filename, xml_filename,
                     author_prefix=None, count=None, publisher=None, shops=None,
                     available=None):
    file_handler = gzopen(csv_filename, "r")
    try:
        csv_iterator = reader(file_handler)

        # great, we have CSV iterator right now which will read our
        # file line by line now let's convert it to stream
        stream = Stream(csv_iterator)

        # now let's fetch author names. Since every row looks like a
        # tuple of (key, value) where key is an author_id and value is
        # a book name we can do key_mapping here. And let's do it in
        # parallel it is I/O bound
        stream = stream.key_map(author_fetch, parallel=True)

        # okay, now let's keep only author name here
        stream = stream.key_map(lambda author: author["name"])

        # we have author prefix, right?
        if author_prefix is not None:
            stream = stream.filter(lambda (author, book): author.startswith(author_prefix))

        # let's fetch publisher now. Let's do it in 10 threads
        if publisher is not None:
            stream = stream.map(
                lambda (author, book): (author, book, publisher_fetch(author, book)),
                parallel=10
            )
            stream = stream.filter(lambda item: item[-1] == publisher)
            # we do not have to have publisher now, let's remove it
            stream = stream.map(lambda item: item[:2])

        # good. Let's compose the list of shops here
        stream.map(
            lambda (author, book): (author, book, shop_prices_fetch(author, book, shops))
        )

        # now let's make averages
        stream.map(lambda item: item[:2] + sum(item[3]) / len(item[3]))

        # let's remove unavailable books now.
        if available is not None:
            if available:
                stream = stream.filter(lambda item: item[-1])
            else:
                stream = stream.filter(lambda item: not item[-1])

        # ok, great. Now we have only those entries which we are requiring
        # let's compose xml now. Remember whole our data won't fit in memory.
        with open(xml_filename, "w") as xml:
            xml.write("<?xml version='1.0' encoding='UTF-8' standalone='yes'?>\n")
            xml.write("<books>\n")
            for author, book, average in stream:
                book_element = ET.Element("book")
                ET.SubElement(book_element, "name").text = unicode(book)
                ET.SubElement(book_element, "author").text = unicode(author)
                ET.SubElement(book_element, "average_price").text = unicode(average)
                xml.write(ET.dumps(book_element) + "\n")
            xml.write("</books>\n")
    finally:
        file_handler.close()

That’s it. On every step we’ve manipulated with given stream to direct it in the way we need. We’ve parallelized where neccessary and actually nothing was executed before we started to iterate the stream. Stream is lazy and it yields one record by one so we haven’t swaped.

I guess it is a time to proceed to API documentation. Actually you need to check only Stream class methods documentation, the rest of are utility ones.

Streams API

This chapter contains documentation on Streams API. As a rule you have to use documentation on Stream class only but if you want you can check internals also.

streams module contains just a Stream class. Basically you want to use only this class and nothing else from the module.

class streams.Stream(iterator, max_cache=0)[source]

Stream class provides you with the basic functionality of Streams. Please checkout member documentation to get an examples.

__init__(iterator, max_cache=0)[source]

Initializes the Stream.

Actually it does some smart handling of iterator. If you give it an instance of dict or its derivatives (such as collections.OrderedDict), it will iterate through it’s items (key and values). Otherwise just normal iterator would be used.

Parameters:
  • iterator (Iterable) – Iterator which has to be converted into Stream.
  • max_cache (int) – the number of items to cache (defaults to Stream.ALL).
__iter__()[source]

To support iteration protocol.

__len__()[source]

To support len() function if given iterator supports it.

__reversed__()[source]

To support reversed() iterator.

all(predicate=<type 'bool'>, **concurrency_kwargs)[source]

Check if all elements matching given predicate exist in the stream. If predicate is not defined, bool() is used.

Parameters:
  • predicate (function) – Predicate to apply to each element of the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords as for Stream.map().
Returns:

The result if we have matched elements or not.

>>> stream = Stream.range(5)
>>> stream.all(lambda item: item > 100)
... False
any(predicate=<type 'bool'>, **concurrency_kwargs)[source]

Check if any element matching given predicate exists in the stream. If predicate is not defined, bool() is used.

Parameters:
  • predicate (function) – Predicate to apply to each element of the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords as for Stream.map().
Returns:

The result if we have matched elements or not.

>>> stream = Stream.range(5)
>>> stream.any(lambda item: item < 100)
... True
average()[source]

Calculates the average of elements in the stream.

Returns:The average of elements.
>>> stream = Stream.range(10000)
>>> stream.average()
... 4999.5
cache(max_cache=<object object>)[source]

Return a stream which caches elements for future iteration.

By default the new stream will cache all elements. If passing an integer to max_cache, the new stream will cache up to that many of the most recently iterated elements.

Parameters:max_cache (int) – the number of items to cache (defaults to Stream.ALL).
Returns:new processed Stream instance.
>>> stream = Stream.range(10).cache()
>>> list(stream)
... [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
>>> list(stream)
... [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
>>> stream = stream.cache(5)
>>> list(stream)
... [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
>>> list(stream)
... [5, 6, 7, 8, 9]
chain()[source]

If elements of the stream are iterable, tries to flat that stream.

Returns:new processed Stream instance.
>>> stream = Stream.range(3)
>>> stream = stream.tuplify()
>>> stream = stream.chain()
>>> list(stream)
>>> [0, 0, 1, 1, 2, 2]
classmethod concat(*streams)[source]

Lazily concatenates several stream into one. The same as Java 8 concat.

Parameters:streams – The Stream instances you want to concatenate.
Returns:new processed Stream instance.
>>> stream1 = Stream(range(2))
>>> stream2 = Stream(["2", "3", "4"])
>>> stream3 = Stream([list(), dict()])
>>> concatenated_stream = Stream.concat(stream1, stream2, stream3)
>>> list(concatenated_stream)
... [0, 1, "2", "3", "4", [], {}]
count(element=<object object>)[source]

Returns the number of elements in the stream. If element is set, returns the count of particular element in the stream.

Parameters:element (object) – The element we need to count in the stream
Returns:The number of elements of the count of particular element.
decimals()[source]

Tries to convert everything to decimal.Decimal and keeps only successful attempts.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2.0, "3", "4.0", None, {}])
>>> stream = stream.longs()
>>> list(stream)
... [Decimal('1'), Decimal('2'), Decimal('3'), Decimal('4.0')]

Note

It is not the same as stream.map(Decimal) because it removes failed attempts.

Note

It tries to use cdecimal module if possible.

distinct()[source]

Removes duplicates from the stream.

Returns:new processed Stream instance.

Note

All objects in the stream have to be hashable (support __hash__()).

Note

Please use it carefully. It returns new Stream but will keep every element in your memory.

divisible_by(number)[source]

Filters stream for the numbers divisible by the given one.

Parameters:number (int) – Number which every element should be divisible by.
Returns:new processed Stream instance.
>>> stream = Stream.range(6)
>>> stream = stream.divisible_by(2)
>>> list(stream)
... [0, 2, 4]
evens()[source]

Filters and keeps only even numbers from the stream.

Returns:new processed Stream instance.
>>> stream = Stream.range(6)
>>> stream = stream.evens()
>>> list(stream)
... [0, 2, 4]
exclude(predicate, **concurrency_kwargs)[source]

Excludes items from Stream according to the predicate. You can consider behaviour as the same as for itertools.ifilterfalse().

As Stream.filter() it also supports parallelization. Please checkout Stream.map() keyword arguments.

Parameters:
  • predicate (function) – Predicate for filtering elements of the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords as for Stream.map().
Returns:

new processed Stream instance.

>>> stream = Stream.range(6)
>>> stream = stream.exclude(lambda item: item % 2 == 0)
>>> list(stream)
... [1, 3, 5]
exclude_nones()[source]

Excludes None from the stream.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, None, 3, None, 4])
>>> stream = stream.exclude_nones()
>>> list(stream)
... [1, 2, 3, 4]
filter(predicate, **concurrency_kwargs)[source]

Does filtering according to the given predicate function. Also it supports parallelization (if predicate is pretty heavy function).

You may consider it as equivalent of itertools.ifilter() but for stream with a possibility to parallelize this process.

Parameters:
  • predicate (function) – Predicate for filtering elements of the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords as for Stream.map().
Returns:

new processed Stream instance.

>>> stream = Stream.range(5)
>>> stream = stream.filter(lambda item: item % 2 == 0)
>>> list(stream)
... [0, 2, 4]
first

Returns a first element from iterator and does not changes internals.

>>> stream = Stream.range(10)
>>> stream.first
... 0
>>> stream.first
... 0
>>> list(stream)
... [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
floats()[source]

Tries to convert everything to float() and keeps only successful attempts.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, "3", "4", None, {}, 5])
>>> stream = stream.floats()
>>> list(stream)
... [1.0, 2.0, 3.0, 4.0, 5.0]

Note

It is not the same as stream.map(float) because it removes failed attempts.

instances_of(cls)[source]

Filters and keeps only instances of the given class.

Parameters:cls (class) – Class for filtering.
Returns:new processed Stream instance.
>>> int_stream = Stream.range(4)
>>> str_stream = Stream.range(4).strings()
>>> result_stream = Stream.concat(int_stream, str_stream)
>>> result_stream = result_stream.instances_of(str)
>>> list(result_stream)
... ['0', '1',  '2', '3']
ints()[source]

Tries to convert everything to int() and keeps only successful attempts.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, "3", "4", None, {}, 5])
>>> stream = stream.ints()
>>> list(stream)
... [1, 2, 3, 4, 5]

Note

It is not the same as stream.map(int) because it removes failed attempts.

classmethod iterate(function, seed_value)[source]

Returns seed stream. The same as for Java 8 iterate.

Returns an infinite sequential ordered Stream produced by iterative application of a function f to an initial element seed, producing a Stream consisting of seed, f(seed), f(f(seed)), etc.

The first element (position 0) in the Stream will be the provided seed. For n > 0, the element at position n, will be the result of applying the function f to the element at position n - 1.

Parameters:
  • function (function) – The function to apply to the seed.
  • seed_value (object) – The seed value of the function.
Returns:

new processed Stream instance.

>>> stream = Stream.iterate(lambda value: value ** 2, 2)
>>> iterator = iter(stream)
>>> next(iterator)
... 2
>>> next(iterator)
... 4
>>> next(iterator)
... 8
key_map(predicate, **concurrency_kwargs)[source]

Maps only key in (key, value) pair. If element is single one, then it would be Stream.tuplify() first.

Parameters:
  • predicate (function) – Predicate to apply to the key of element in the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords as for Stream.map().
Returns:

new processed Stream instance.

>>> stream = Stream.range(4)
>>> stream = stream.tuplify()
>>> stream = stream.key_map(lambda item: item ** 3)
>>> list(stream)
... [(0, 0), (1, 1), (8, 2), (27, 3)]
>>> stream = Stream.range(4)
>>> stream = stream.key_map(lambda item: item ** 3)
>>> list(stream)
... [(0, 0), (1, 1), (8, 2), (27, 3)]
keys()[source]

Iterates only keys from the stream (first element from the tuple). If element is single then it will be used.

Returns:new processed Stream instance.
>>> stream = Stream.range(5)
>>> stream = stream.key_map(lambda item: item ** 3)
>>> stream = stream.keys()
>>> list(stream)
... [0, 1, 8, 27, 64]
largest(size)[source]

Returns size largest elements from the stream.

Returns:new processed Stream instance.
>>> stream = Stream.range(3000)
>>> stream.largest(5)
>>> list(stream)
>>> [2999, 2998, 2997, 2996, 2995]
limit(size)[source]

Limits stream to given size.

Parameters:size (int) – The size of new Stream.
Returns:new processed Stream instance.
>>> stream = Stream.range(1000)
>>> stream = stream.limit(5)
>>> list(stream)
... [0, 1, 2, 3, 4]
longs()[source]

Tries to convert everything to long() and keeps only successful attempts.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, "3", "4", None, {}, 5])
>>> stream = stream.longs()
>>> list(stream)
... [1L, 2L, 3L, 4L, 5L]

Note

It is not the same as stream.map(long) because it removes failed attempts.

map(predicate, **concurrency_kwargs)[source]

The corner method of the Stream and others are basing on it. It supports parallelization out of box. Actually it works just like itertools.imap().

Parameters:
  • predicate (function) – Predicate to map each element of the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords.
Returns:

new processed Stream instance.

Parallelization is configurable by keywords. There is 2 keywords supported: parallel and process. If you set one keyword to True then Stream would try to map everything concurrently. If you want more intelligent tuning just set the number of workers you want.

For example, you have a list of URLs to fetch

>>> stream = Stream(urls)

You can fetch them in parallel

>>> stream.map(requests.get, parallel=True)

By default, the number of workers is the number of cores on your computer. But if you want to have 64 workers, you are free to do it

>>> stream.map(requests.get, parallel=64)

The same for process which will try to use processes.

>>> stream.map(requests.get, process=True)

and

>>> stream.map(requests.get, process=64)

Note

Python multiprocessing has its caveats and pitfalls, please use it carefully (especially predicate). Read the documentation on multiprocessing and try to google best practices.

Note

If you set both parallel and process keywords only parallel would be used. If you want to disable some type of concurrency just set it to None.

>>> stream.map(requests.get, parallel=None, process=64)

is equal to

>>> stream.map(requests.get, process=64)

The same for parallel

>>> stream.map(requests.get, parallel=True, process=None)

is equal to

>>> stream.map(requests.get, parallel=True)

Note

By default no concurrency is used.

median()[source]

Returns median value from the stream.

Returns:The median of the stream.
>>> stream = Stream.range(10000)
>>> stream.median()
... 5000

Note

Please be noticed that all elements from the stream would be fetched in the memory.

nth(nth_element)[source]

Returns Nth element from the stream.

Parameters:nth_element (int) – Number of element to return.
Returns:Nth element.
>>> stream = Stream.range(10000)
>>> stream.average()
... 4999.5

Note

Please be noticed that all elements from the stream would be fetched in the memory (except of the case where nth_element == 1).

odds()[source]

Filters and keeps only odd numbers from the stream.

Returns:new processed Stream instance.
>>> stream = Stream.range(6)
>>> stream = stream.odds()
>>> list(stream)
... [1, 3, 5]
only_falses()[source]

Keeps only those elements where bool(item) == False.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, None, 0, {}, [], 3])
>>> stream = stream.only_trues()
>>> list(stream)
... [None, 0, {}, []]

Opposite to Stream.only_trues().

only_nones()[source]

Keeps only None in the stream (for example, for counting).

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, None, 3, None, 4])
>>> stream = stream.only_nones()
>>> list(stream)
... [None, None]
only_trues()[source]

Keeps only those elements where bool(element) == True.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2, None, 0, {}, [], 3])
>>> stream = stream.only_trues()
>>> list(stream)
... [1, 2, 3]
partly_distinct()[source]

Excludes some duplicates from the memory.

Returns:new processed Stream instance.

Note

All objects in the stream have to be hashable (support __hash__()).

Note

It won’t guarantee you that all duplicates will be removed especially if your stream is pretty big and cardinallity is huge.

peek(predicate)[source]

Does the same as Java 8 peek.

Parameters:predicate (function) – Predicate to apply on each element.
Returns:new processed Stream instance.

Returns a stream consisting of the elements of this stream, additionally performing the provided action on each element as elements are consumed from the resulting stream.

classmethod range(*args, **kwargs)[source]

Creates numerial iterator. Absoultely the same as Stream.range(10) and Stream(range(10)) (in Python 2: Stream(xrange(10))). All arguments go to range() (xrange()) directly.

Returns:new processed Stream instance.
>>> stream = Stream.range(6)
>>> list(stream)
... [0, 1, 2, 3 ,4, 5]
>>> stream = Stream.range(1, 6)
>>> list(stream)
... [1, 2, 3, 4, 5]
>>> stream = Stream.range(1, 6, 2)
>>> list(stream)
... [1, 3, 5]
reduce(function, initial=<object object>)[source]

Applies reduce() for the iterator

Parameters:
  • function (function) – Reduce function
  • initial (object) – Initial value (if nothing set, first element) would be used.
>>> Stream = stream.range(5)
>>> stream.reduce(operator.add)
... 10
regexp(regexp, flags=0)[source]

Filters stream according to the regular expression using re.match(). It also supports the same flags as re.match().

Parameters:
  • regexp (str) – Regular expression for filtering.
  • flags (int) – Flags from re.
Returns:

new processed Stream instance.

>>> stream = Stream.range(100)
>>> stream = stream.strings()
>>> stream = stream.regexp(r"^1")
>>> list(stream)
... ['1', '10', '11', '12', '13', '14', '15', '16', '17', '18', '19']
reversed()[source]

Reverses the stream.

Returns:new processed Stream instance.
... note::
If underlying iterator won’t support reversing, we are in trouble and need to fetch everything into the memory.
skip(size)[source]

Skips first size elements.

Parameters:size (int) – The amount of elements to skip.
Returns:new processed Stream instance.
>>> stream = Stream.range(10)
>>> stream = stream.skip(5)
>>> list(stream)
... [5, 6, 7, 8, 9]
smallest(size)[source]

Returns size largest elements from the stream.

Returns:new processed Stream instance.
>>> stream = Stream.range(3000)
>>> stream.smallest(5)
>>> list(stream)
>>> [0, 1, 2, 3, 4]
sorted(key=None, reverse=False)[source]

Sorts the stream elements.

Parameters:
  • key (function) – Key function for sorting
  • reverse (bool) – Do we need to sort in descending order?
Returns:

new processed Stream instance.

... note::
Of course no magic here, we need to fetch all elements for sorting into the memory.
strings()[source]

Tries to convert everything to unicode() (str for Python 3) and keeps only successful attempts.

Returns:new processed Stream instance.
>>> stream = Stream([1, 2.0, "3", "4.0", None, {}])
>>> stream = stream.strings()
>>> list(stream)
... ['1', '2.0', '3', '4.0', 'None', '{}']

Note

It is not the same as stream.map(str) because it removes failed attempts.

Note

It tries to convert to unicode if possible, not bytes.

sum()[source]

Returns the sum of elements in the stream.

>>> Stream = stream.range(10)
>>> stream = stream.decimals()
>>> stream = stream.sum()
... Decimal('45')

Note

Do not use sum() here. It does sum regarding to defined __add__() of the classes. So it can sum decimal.Decimal with int for example.

tuplify(clones=2)[source]

Tuplifies iterator. Creates a tuple from iterable with clones elements.

Parameters:clones (int) – The count of elements in result tuple.
Returns:new processed Stream instance.
>>> stream = Stream.range(2)
>>> stream = stream.tuplify(3)
>>> list(stream)
... [(0, 0, 0), (1, 1, 1)]
value_map(predicate, **concurrency_kwargs)[source]

Maps only value in (key, value) pair. If element is single one, then it would be Stream.tuplify() first.

Parameters:
  • predicate (function) – Predicate to apply to the value of element in the Stream.
  • concurrency_kwargs (dict) – The same concurrency keywords as for Stream.map().
Returns:

new processed Stream instance.

>>> stream = Stream.range(4)
>>> stream = stream.tuplify()
>>> stream = stream.value_map(lambda item: item ** 3)
>>> list(stream)
... [(0, 0), (1, 1), (2, 8), (3, 27)]
>>> stream = Stream.range(4)
>>> stream = stream.value_map(lambda item: item ** 3)
>>> list(stream)
... [(0, 0), (1, 1), (2, 8), (3, 27)]
values()[source]

Iterates only values from the stream (last element from the tuple). If element is single then it will be used.

Returns:new processed Stream instance.
>>> stream = Stream.range(5)
>>> stream = stream.key_map(lambda item: item ** 3)
>>> stream = stream.values()
>>> list(stream)
... [0, 1, 2, 3, 4]

Internal modules

Basically you do not need this API but if you are curious feel free to check it out.

streams.executors

This module provides different implementation of concurrent executors suitable to work with streams.poolofpools.PoolOfPools. If Gevent is available then you can also import streams.executors._gevent.GeventExecutor here.

Also it has some class called streams.executors.ParallelExecutor. This is dynamically calculated class for default concurrent execution. If code is monkey patched by Gevent, then it uses streams.executors._gevent.GeventExecutor. Otherwise - streams.executors.executors.ThreadPoolExecutor.

streams.executors.executors

This module has implementation of executors wrapped by streams.executors.mixins.PoolOfPoolsMixin and applicable to work with streams.poolofpools.PoolOfPools.

Basically all of them are thin extensions of classes from concurrent.futures.

class streams.executors.executors.ProcessPoolExecutor(max_workers=None)[source]

Implementation of concurrent.futures.ProcessPoolExecutor applicable to work with streams.poolofpools.PoolOfPools.

class streams.executors.executors.SequentalExecutor(*args, **kwargs)[source]

Debug executor. No concurrency, it just yields elements one by one.

class streams.executors.executors.ThreadPoolExecutor(max_workers)[source]

Implementation of concurrent.futures.ThreadPoolExecutor applicable to work with streams.poolofpools.PoolOfPools.

streams.executors.mixins

This module provides PoolOfPoolMixin only. Basically you need to mix it into concurrent.futures.Executor implementation and it will be possible to use it with PoolOfPools.

class streams.executors.mixins.PoolOfPoolsMixin[source]

Mixin to support streams.poolofpools.PoolOfPools execution properly.

Basically it replaces map implementation and provides some additional interface which helps streams.poolofpools.PoolOfPools to manage executor instance. Current implementation supports expanding only (dynamically increasing, on the fly) the number of workers.

static dummy_callback(*args, **kwargs)[source]

Just a dummy callback if no streams.poolofpools.PoolOfPools.worker_finished() is supplied for the mapper. Basically does nothing. Literally nothing. Good thing though, no bugs.

expand(expand_to)[source]

The hack to increase an amount of workers in executor.

Parameters:expand_to (int) – The amount of worker we need to add to the executor.

Note

It works perfect with streams.executors._gevent.GeventExecutor and concurrent.futures.ThreadPoolExecutor but has some issues with concurrent.futures.ProcessPoolExecutor.

It increases the amount of workers who manage task queue but it is not possible to expand queue itself in a good way (current implementation has a limit of tasks in the queue).

static get_first(queue)[source]

Extracts the result of the execution from the first element of the queue (to support order since a map is ordering function). Also it tries to handle exceptions if presented in the same way as concurrent.futures.ThreadPoolExecutor or concurrent.futures.ProcessPoolExecutor do.

Note

It relies on given implementation of map method in both concurrent.futures.ThreadPoolExecutor and concurrent.futures.ProcessPoolExecutor so if you see some differences in behaviour please create an issue.

map(fn, *iterables, **kwargs)[source]

New implementation of concurrent mapper.

It has 2 new arguments: callback and required_workers

Parameters:
  • callback (Callable) – Callback to execute after map is done
  • required_workers (int) – The amount of workers we have to use for this map procedure.
It differs from default implementation in 2 ways:
  1. It uses the limit of workers (required_workers). It can be less than max workers defined on executor initialization hence it is possible to utilize the same executor for several tasks more efficient.
  2. It doesn’t create a list of futures in memory. Actually it creates only required_workers amount of futures and tries to keep this count the same during whole procedure. Yes, it is not naturally concurrent execution because it just submits task by task but on big iterables it utilizes as less memory as possible providing reasonable concurrency.
streams.executors._gevent

This module provides implementation of streams.executors._gevent.GreenletFuture (thin wrapper around concurrent.futures.Future) and implementation of streams.executors._gevent.GeventExecutor.

Basically you can use concurrent.futures.ThreadPoolExecutor, it is ok and will work but to utilize the power of greenlets more carefully it makes sense to use custom one.

class streams.executors._gevent.GeventExecutor(*args, **kwargs)[source]

Implementation of Gevent executor fully compatible with concurrent.futures.Executor.

class streams.executors._gevent.GreenletFuture(greenlet)[source]

Just a thin wrapper around a concurrent.futures.Future to support greenlets.

streams.iterators

This module contains some useful iterators. Consider it as a small ad-hoc extension pack for itertools.

streams.iterators.accumulate(iterable, function=<built-in function add>)[source]

Implementation of itertools.accumulate() from Python 3.3.

streams.iterators.distinct(iterable)[source]

Filters items from iterable and returns only distinct ones. Keeps order.

Parameters:iterable (Iterable) – Something iterable we have to filter.
>>> list(distinct([1, 2, 3, 2, 1, 2, 3, 4]))
... [1, 2, 3, 4]

Note

This is fair implementation and we have to keep all items in memory.

Note

All items have to be hashable.

streams.iterators.partly_distinct(iterable)[source]

Filters items from iterable and tries to return only distincts. Keeps order.

Parameters:iterable (Iterable) – Something iterable we have to filter.
>>> list(partly_distinct([1, 2, 3, 2, 1, 2, 3, 4]))
... [1, 2, 3, 4]

Note

Unlike distinct() it won’t guarantee that all elements would be distinct. But if you have rather small cardinality of the stream, this would work.

Note

Current implementation guarantees support for 10000 distinct values. If your cardinality is bigger, there might be some duplicates.

streams.iterators.peek(iterable, function)[source]

Does the same as Java 8 peek does.

Parameters:
  • iterable (Iterable) – Iterable we want to peek
  • function (function) – Peek function
>>> def peek_func(item):
...     print "peek", item
>>> list(peek([1, 2, 3], peek_func))
... peek 1
... peek 2
... peek 3
... [1, 2, 3]
streams.iterators.seed(function, seed_value)[source]

Does the same as Java 8 iterate.

Parameters:
  • iterable (Iterable) – Iterable we want to peek
  • function (function) – Peek function
>>> iterator = seed(lambda x: x * 10, 1)
>>> next(iterator)
... 1
>>> next(iterator)
... 10
>>> next(iterator)
... 100

streams.poolofpools

class streams.poolofpools.ExecutorPool(worker_class)[source]

Executor pool for PoolOfPools which does accurate and intelligent management for the pools of predefined classes.

Basically it tries to reuse existing executors if possible. If it is not possible it creates new ones.

Just an example: you’ve done a big mapping of the data in 10 threads. As a rule you need to shutdown and clean this pool. But a bit later you see that you need for the pool of 4 threads. Why not to reuse existing pool? This class allow you to do that and it tracks that 6 threads are idle. So if you will have a task where you need <= 6 threads it will reuse that pool also. Task with 4 threads may continue to work in parallel but you have 6 threads you can occupy. So this is the main idea.

Also it tries to squash pools into single instance if you have several which idle by expanding an amount of workers in one instance throwing out another one.

__init__(worker_class)[source]

Constructor of the class. worker_class has to be a class which supports required interface and behaviour, it has to be an instance of streams.executors.mixins.PoolOfPoolsMixin.

Parameters:worker_class (PoolOfPoolsMixin) – The class of executors this pool has to maintain.
__weakref__

list of weak references to the object (if defined)

get(required_workers)[source]

Returns a mapper which guarantees that you can utilize given number of workers.

Parameters:required_workers (int) – The number of workers you need to utilize for your task.
get_any()[source]

Returns any map function, it is undetermined how many workers does it have. As a rule, you get a minimal amount of workers within a pool of executors.

get_suitable_worker(required_workers)[source]

Returns suitable executor which has required amount of workers. Returns None if nothing is available.

Actually it returns a tuple of worker and a count of workers available for utilization within a given pool. It may be more than required_workers but it can’t be less.

Parameters:required_workers (int) – The amount of workers user requires.
name_to_worker_mapping()[source]

Maps worker names (the result of applying id() to the executor) to executor instances.

real_worker_availability()[source]

Returns mapping of the name for the executor and it real availability. Since worker_finished() does not do any defragmentation of availability it may be possible that internal structure contains multiple controversial information about worker availability. This method is intended to restore the truth.

squash()[source]

Squashes pools and tries to minimize the amount of pools available to avoid unnecessary fragmentation and complexity.

squash_workers(names, avails)[source]

Does actual squashing/defragmentation of internal structure.

worker_finished(worker, required_workers)[source]

The callback used by streams.executors.mixins.PoolOfPoolsMixin.

class streams.poolofpools.PoolOfPools[source]

Just a convenient interface to the set of multiple ExecutorPool instances, nothing more.

__weakref__

list of weak references to the object (if defined)

get(kwargs)[source]

Returns the mapper.

Parameters:kwargs (dict) – Keyword arguments for the mapper. Please checkout streams.Stream.map() documentation to understand what this dict has to have.
static get_from_pool(pool, required_workers)[source]

Fetches mapper from the pool.

Parameters:
  • pool (ExecutorPool) – The pool you want to fetch mapper from.
  • required_workers (int) – The amount of workers you are requiring. It can be None then ExecutorPool.get_any() would be executed.
parallel(required_workers)[source]

Fetches parallel executor mapper from the underlying ExecutorPool.

Parameters:required_workers (int) – The amount of workers you are requiring. It can be None then ExecutorPool.get_any() would be executed.
process(required_workers)[source]

Fetches process executor mapper from the underlying ExecutorPool.

Parameters:required_workers (int) – The amount of workers you are requiring. It can be None then ExecutorPool.get_any() would be executed.

streams.utils

This module contains some utility functions for Streams.

You may wonder why do we need for such simple filter-* functions. The reason is simple and this is about how multiprocessing and therefore concurrent.futures.ProcessPoolExecutor works. It can’t pickle lambdas so we need for whole pickleable functions.

class streams.utils.MaxHeapItem(value)[source]

This is small wrapper around item to give it a possibility to use heaps from heapq as max-heaps. Unfortunately this module provides min-heaps only.

Guys, come on. We need for max-heaps to.

streams.utils.apply_to_tuple(*funcs, **kwargs)[source]

Applies several functions to one item and returns tuple of results.

Parameters:
  • func (list) – The list of functions we need to apply.
  • kwargs (dict) – Keyword arguments with only one mandatory argument, item. Functions would be applied to this item.
>>> apply_to_tuple(int, float, item="1")
... (1, 1.0)
streams.utils.decimal_or_none(item)[source]

Tries to convert item to decimal.Decimal. If it is not possible, returns None.

Parameters:item (object) – Element to convert into decimal.Decimal.
>>> decimal_or_none(1)
... Decimal("1")
>>> decimal_or_none("1")
... Decimal("1")
>>> decimal_or_none("smth")
... None
streams.utils.filter_false(argument)[source]

Opposite to streams.utils.filter_true()

Parameters:argument (tuple) – Argument consists of predicate function and item iteself.
>>> filter_false((lambda x: x <= 5, 5))
... False, 5
>>> filter_false((lambda x: x > 100, 1))
... True, 1
streams.utils.filter_keys(item)[source]

Returns first element of the tuple or item itself.

Parameters:item (object) – It can be tuple, list or just an object.
>>> filter_keys(1)
... 1
>>> filter_keys((1, 2))
... 1
streams.utils.filter_true(argument)[source]

Return the predicate value of given item and the item itself.

Parameters:argument (tuple) – Argument consists of predicate function and item iteself.
>>> filter_true((lambda x: x <= 5, 5))
... True, 5
>>> filter_true((lambda x: x > 100, 1)
... False, 1
streams.utils.filter_values(item)[source]

Returns last element of the tuple or item itself.

Parameters:item (object) – It can be tuple, list or just an object.
>>> filter_values(1)
... 1
>>> filter_values((1, 2))
... 2
streams.utils.float_or_none(item)[source]

Tries to convert item to float(). If it is not possible, returns None.

Parameters:item (object) – Element to convert into float().
>>> float_or_none(1)
... 1.0
>>> float_or_none("1")
... 1.0
>>> float_or_none("smth")
... None
streams.utils.int_or_none(item)[source]

Tries to convert item to int(). If it is not possible, returns None.

Parameters:item (object) – Element to convert into int().
>>> int_or_none(1)
... 1
>>> int_or_none("1")
... 1
>>> int_or_none("smth")
... None
streams.utils.key_mapper(argument)[source]

Maps predicate only to key (first element) of a item. If item is not tuple() then tuplifies it first.

Parameters:argument (tuple) – The tuple of (predicate and item).
>>> key_mapper((lambda x: x + 10, (1, 2)))
... (11, 2)
streams.utils.long_or_none(item)[source]

Tries to convert item to long(). If it is not possible, returns None.

Parameters:item (object) – Element to convert into long().
>>> long_or_none(1)
... 1L
>>> long_or_none("1")
... 1L
>>> long_or_none("smth")
... None
streams.utils.make_list(iterable)[source]

Makes a list from given iterable. But won’t create new one if iterable is a list() or tuple() itself.

Parameters:iterable (Iterable) – Some iterable entity we need to convert into list().
streams.utils.unicode_or_none(item)[source]

Tries to convert item to unicode(). If it is not possible, returns None.

Parameters:item (object) – Element to convert into unicode().
>>> unicode_or_none(1)
... u"1"
>>> unicode_or_none("1")
... u"1"
>>> unicode_or_none("smth")
... u"smth"

Note

This is relevant for Python 2 only. Python 3 will use native str().

streams.utils.value_mapper(argument)[source]

Maps predicate only to value (last element) of a item. If item is not tuple() then tuplifies it first.

Parameters:argument (tuple) – The tuple of (predicate and item).
>>> value_mapper((lambda x: x + 10, (1, 2)))
... (1, 12)

Indices and tables