collection-pipelines 0.1.3

Framework to implement collection pipelines in python.

Experimental collection pipeline pattern implementation in python.

cat('/tmp/file.txt') | filter('some line') | filter('some line 2') | out()

• Usage

• Reference

  • Sources

  • Transformers, Filters

  • Output Processors

• Development

  • Sources

  • Transformers, Filters

  • Output processors

• More Samples

  • Bar Chart

  • Line Chart

  • Word Cloud

Note

Library only works on Python 3. There are no plans to support previous python versions.

Usage

Every pipeline has a data source (cat(), http(), etc.) and an optional data transformation/filtering and output processors.

     Source        Transformations      Output
+---------------+----------------------+-------+
|               |                      |       |
v               v                      v       V
 echo('1.2.3.4') | split('.') | count() | out()

You can save a partial pipe and reuse it later.

from collection_pipelines import *

word_list = echo('word1 word2 word3') | words()
word_list | out() # will print the words to stdout

word_list | filter(word2) | freq() | bar() # will draw a bar chart for word frequencies

Reference

Sources

cat(file_name)

Reads the specified file. Sends items to the pipe line by line.

echo(text)

Sends the specified text to the pipe.

http(url)

Sends HTTP GET method to the specified URL and puts the response body to pipe.

Transformers, Filters

filter(value)

Filters out the items that match the specified value.

head(N)

Passes only the first N items through the pipe.

tail(N)

Passes only the last N items through the pipe.

count()

Calculates the incoming items. When the pipeline source signals the end of the items, count() sends a single item - numbers of items, to the pipe.

freq()

Calculates how many times each unique item appears on the pipe. When the source signals the end of the items, every unique item is sent together with it’s repetition count. freq() outputs tuples: (‘item_x’, 8).

unique()

Filters out the duplicate items.

json(path)

Parses incoming items as json strings, extracts elements with the specified json path and sends them trough the pipe. E.g. let’s say we have json ‘{“name”: {“first”: “Bob”}}’ and we want to extract the first name. Then path would be “name.first”.

split(delimiter)

Splits the incoming items by the specified delimiter. Sends the resulting items through the pipeline one by one.

words(text)

Splits the incoming text items into words and sends each word through the pipe.

Output Processors

out()

Outputs items to stdandard output.

value()

Returns the collected items rather than outputting them somewhere. Useful when you want to store resulting pipeline items to variable. If more than one item passes the pipeline, the array of those items is returned.

line()

Collects all items and draws a line chart. Items must be tuples where first item is X axis value, and second item is Y axis value. Chart is plotted using matplotlib.

bar()

Collects all items and draws a bar chart. Items must be tuples where first item is X axis value, and second item is Y axis value. Chart is plotted using matplotlib.

wordcloud()

Collects all text items and draws a word cloud. See: https://github.com/amueller/word_cloud

Development

If you want to write your own sources, transformers or outputs there’s couple of base classes you should get familiar with.

Let’s implement a very basic filter that forwards only even numbers.

from collection_pipelines import *

class even(CollectionPipelineProcessor):
    def process(self, item):
        if isinstance(item, int):
            if item % 2 == 0:
                self.receiver.send(item)

echo([1, 2, 3]) | even() | out()

Sources

Every source object must extend the CollectionPipelineSource class and implement the on_begin() method.

E.g. this source will send random integer to a pipeline:

import random

class rand_int(CollectionPipelineSource):
    def on_begin(self):
        self.receiver.send(random.randint(0, 1000))
        self.receiver.close()

Transformers, Filters

Every transformer and filter is a python object that instantiates a class that extends CollectionPipelineProcessor class. All the work is done in process() method. This methods receives an item passing the pipeline.

You might either ignore, transform or simply pass forward the items. To send item further to the pipe use self.receiver.send(item).

E.g. if you wanted to multiply all items, you could implement the method like this

def process(self, item):
    self.receiver.send(item * 2)

Output processors

Pipeline output processors must extend the CollectionPipelineOutput class. Output processors are special in a way that they don’t forwards the items any further. They trigger the pipeline execution.

Implementing an output processor is very similar to implementing a transformer.

class stdout(CollectionPipelineOutput):
    def process(self, item):
        print(item)

Such processor would print an item as soon as it received one. There’s also a special method on_done(), which is called when all items in the pipeline are processed.

E.g. if you wanted an output processor to print items only when you received all of them, the class would look like

class stdout(CollectionPipelineOutput):
    def __init__(self):
        self.items = []

    def process(self, item):
        self.items.append(item)

    def on_done(self):
        for item in self.items:
            print(item)

More Samples

Bar Chart

echo([('apples', 2), ('bananas', 5), ('oranges', 3)]) | bar()

Line Chart

echo([(1, 10), (2, 7), (3, 5), (4, 5), (5, 8)]) | line()

Word Cloud

cat('README.rst') | wordcloud()