datagrab 0.1.4

The easy way to access and interpret textual web resources

datagrab - the easy way to access and interpret textual web resources!

Overview

Let's be clear: requests is an awesome library. But even so, some really basic use cases still seem to require a lot of lines of code!

datagrab was written as a set of classes and methods designed to simplify using the requests library for some typical web scraping and REST API use cases. In short: download, parse and process the returned text easily, so you are closer to doing something useful with it.

• Request data and ensure the desired result arrived with a single method (retrieve_response module)
• Connect to REST API services with Basic Authentication implemented for you (RESTConnect module)
• Parse the data and use simple functions to pull out the desired data (interpret_response module)

Usage

Getting it

To download datagrab, either fork this github repo or simply use Pypi via pip.

Using it

Basic usage

Getting a response and some text values

Let's assume the server is not expecting any specific header or other content:

# RetrievedResponse is the class used to get the raw http response
>>> from datagrab.retrieve_response.retrieve_response import RetrievedResponse
>>> from datagrab.interpret_response.interpret_html_response import (
  ResponseInterpreter)

>>> rr = RetrievedResponse("https://www.bbc.co.uk")

# Send the http request and ensure that we actually got a 200 response
# This method incorporates a number of exception handlers for most types
# of http response codes (too many redirects, server not found etc.)

>>> rr.getValidate()

>>> rr.response #attribute created by getValidate method just executed
<Response [200]>

>>> ri = ResponseInterpreter(rr.response)

# print anything with a h1-tag
>>> for i in ri.getTextByElementType("h1"):
      print(i)

BBC Homepage

Note that getTextByElementType returns a object. This is handy because you probably need an iterable in any case.

Getting attribute text

Let's say you want to retrieve all the link urls on a page (a common toy example). Then you'd need all the hrefs of all <a> tags.

# Continued from above
>>> hrefs = ri.getAttributeText("a", "href")
>>> next(hrefs)
"https://www.bbc.co.uk"

Searching by attribute

We cover two use cases here:

1. We want to find elements which have a specific attribute, where that attribute can take any value.
2. We want to find elements with an attribute having a specific value.

Let's see it in action.

# Continued from above
# Search for anything with a src attribute

>>> srcs = ri.getElementsByType(True, # search for anything with a truthy tag
                                    # - i.e. any tag
                attrs="src")       # with a src attribute

>>> srcs[0] # getElementsByType method returns a list, so we can index it normally

<script src=*bbc script file*><script>

# Search for any link sending the user to the Homepage
>>> hp_links = ri.getElementsByType("a", attrs={"href":"https://www.bbc.co.uk"})
>>> hp_links[0]

<a href="https://www.bbc.co.uk">Homepage</a>

It's worth noting that the elements we retrieve here are still just BeautifulSoup element.Tag instances. So you can still access properties like .text, .attrs etc. if you find that more intuitive.

Adding support for access to child nodes as a method of the ResponseInterpreter class is a #TODO, but for now you can use the BeautifulSoup methods explained in this StackOverflow question.

Intermediate usage

Connecting to REST services

If the service you want to get resources from just requires you to submit a url, you do as above up to rr.getValidate().

But if you need Basic Auth, we've got you covered with the following...

>>> from datagrab.RESTConnect.basicAuth import BasicAuth
>>> from datagrab.retrieve_response.retrieve_response import RetrievedResponse

>>> ba = basicAuth(<my_username>, key=<my_app_key>) # Has attribute .basicAuthHeader

Note that the basicAuth class does not need a key kwarg to instantiate or be used successfully. Some REST API's provide non-sensitive data and just require you to encode your app key as the basic auth username. UK's Companies House API at time of writing is a good example of this.

Now, we can use the .basicAuthHeader attribute of our ba instance to enhance the RetrievedResponse class.

# Continued from above
# In keyword argument request_kwargs you can, if needed, add other keyword
# arguments to a requests.get call such as proxies or params.

>>> rr = RetrievedResponse(<my_url>, request_kwargs={"headers":ba.basicAuthHeader})
>>> rr.getValidate()
>>> rr.response
<Response [200]>

And that's it!

OAuth support is a #TODO

Interpreting JSON data

Most RESTful services these days return data in JSON format -- unless you're using steam-powered enterprise ERP's and the like, in which case you're still using xml (for which the interpret_html_response examples above should do the job).

from datagrab.interpret_response.interpret_json_response import JsonResponseInterpreter

This example draws on Brian Dew's brilliant example of using the IMF's data api. This query delivers monthly import price index data for the UK between 2010 and 2011.

# Build the url and request the data
>>> url_base = 'http://dataservices.imf.org/REST/SDMX_JSON.svc/'
>>> query = 'CompactData/IFS/M.GB.PMP_IX?startPeriod=2010&endPeriod=2011' # adjust codes here
>>> query_url = url_base+query
>>> rr = RetrievedResponse(query_url)
>>> rr.getValidate()

# instantiate JsonResponseInterpreter
>>> jri = JsonResponseInterpreter(rr.response)
# If you want, you can look at the raw text of the Response
>>> jri.requestResponseText[:71]
'{"CompactData":{"@xmlns:xsi":"http://www.w3.org/2001/XMLSchema-instance'

One of the more cumbersome parts of working with JSON data is acttually getting to the node that you're interested in. This requires a lot of square brackets, in our case

jri.jsonDict["CompactData"]["DataSet"]["Series"]["Obs"]

Firstly, how do we know that this is the correct series of brackets?

The "easiest" way of getting to grips with a JSON data structure is normally to dump it to a json file and use a desktop IDE to explore it. But suppose we're on a work machine which doesn't let us install anything so nice, or we just don't want to go through the hassle of yet another window on our already cluttered desktop.

You'll be pleased to know that our JsonResponseInterpreter has a solution for that, which is based on the treelib library. It allows us to view a hierarchical element tree of all the nodes in the JSON

>>>jri.visualize_json()
CompactData
├── @xmlns
├── @xmlns:xsd
├── @xmlns:xsi
├── @xsi:schemaLocation
├── CompactData
├── DataSet
│   ├── @xmlns
│   └── Series
│       ├── @BASE_YEAR
│       ├── @FREQ
│       ├── @INDICATOR
│       ├── @REF_AREA
│       ├── @TIME_FORMAT
│       ├── @UNIT_MULT
│       └── Obs
└── Header
    ├── DataSetID
    ├── ID
    ├── Prepared
    ├── Receiver
    │   └── @id
    ├── Sender
    │   ├── @id
    │   ├── Contact
    │   │   ├── Telephone
    │   │   └── URI
    │   └── Name
    │       ├── #text
    │       └── @xml:lang
    └── Test

But we still have the following problem:

Writing all of those square brackets is boring, error-prone and non-intuitive.

What we want to do is feed to a function a list representing the node path that we want to traverse. Well, since you asked...

To check out a particular point on the element tree, we can start using a convenience method attached to our jri object: jri.json_tree_traverse

The bit we're interested in is the "DataSet" child nodes.

>>> jri.json_tree_traverse(["CompactData","DataSet"])
{'@xmlns': 'http://dataservices.imf.org/compact/IFS',
'Series': {'@FREQ': 'M',
'@REF_AREA': 'GB',
'@INDICATOR': 'PMP_IX',
'@UNIT_MULT': '0',
'@BASE_YEAR': '2010=100',
'@TIME_FORMAT': 'P1M',
'Obs': [{'@TIME_PERIOD': '2010-01', '@OBS_VALUE': '96.7710371819961'},
{'@TIME_PERIOD': '2010-02', '@OBS_VALUE': '97.5538160469667'},
{'@TIME_PERIOD': '2010-03', '@OBS_VALUE': '100.391389432485'},
...

# So, it seems we need to traverse to the 'Series'->'Obs' node to get the actual data.

>>> import_price_index_data = jri.json_tree_traverse(
      ["CompactData", "DataSet","Series", "Obs"])

>>> import_price_index_data[:3]
[{'@TIME_PERIOD': '2010-01', '@OBS_VALUE': '96.7710371819961'},
 {'@TIME_PERIOD': '2010-02', '@OBS_VALUE': '97.5538160469667'},
 {'@TIME_PERIOD': '2010-03', '@OBS_VALUE': '100.391389432485'}]

So far, so good.

But now, let's say we want to take what we've got and just look at the value for January 2011.

We have a convenience function for this!

>>> from datagrab.interpret_response.interpret_json_response import (
 query_json_with_func, query_json)

# query_json allows you to query based on the key-value pair

>>> jan_2010 = query_json(import_price_index_data,"@TIME_PERIOD","2011-01")  

Often, you'll actually want to do more sophisticated queries. For example, you might want to see change over time for a specific period of the year.

query_json_with_func is a more flexible option. You can pass it your own filter function.

>>> jan_all = query_json_with_func(import_price_index_data,
                    lambda x: x["@TIME_PERIOD"][-3:]=="-01")

# query_json_with_func returns a `filter` so we'll view it here as a list.

>>> list(jan_all)
[{'@TIME_PERIOD': '2010-01', '@OBS_VALUE': '96.7710371819961'},
 {'@TIME_PERIOD': '2011-01', '@OBS_VALUE': '104.598825831703'}]