Transforms JavaScript objects into Python data structures.
In web scraping, you sometimes need to transform Javascript objects embedded in HTML pages into valid Python dictionaries. chompjs is a library designed to do that as a more powerful replacement of standard json.loads:
>>> chompjs.parse_js_object("{a: 100}")
{'a': 100}
>>>
>>> json_lines = """
... {'a': 12}
... {'b': 13}
... {'c': 14}
... """
>>> for entry in chompjs.parse_js_objects(json_lines):
... print(entry)
...
{'a': 12}
{'b': 13}
{'c': 14}1. installation
> pip install chompjs
or build from source:
$ git clone https://github.com/Nykakin/chompjs
$ cd chompjs
$ python setup.py build
$ python setup.py installThere are two functions available:
parse_js_object- try reading first encountered JSON-like object. RaisesValueErroron failureparse_js_objects- returns a generator yielding all encountered JSON-like objects. Can be used to read JSON Lines. Does not raise on invalid input.
An example usage with scrapy:
import chompjs
import scrapy
class MySpider(scrapy.Spider):
# ...
def parse(self, response):
script_css = 'script:contains("__NEXT_DATA__")::text'
script_pattern = r'__NEXT_DATA__ = (.*);'
# warning: for some pages you need to pass replace_entities=True
# into re_first to have JSON escaped properly
script_text = response.css(script_css).re_first(script_pattern)
try:
json_data = chompjs.parse_js_object(script_text)
except ValueError:
self.log('Failed to extract data from {}'.format(response.url))
return
# work on json_dataParsing of JSON5 objects is supported:
>>> data = """
... {
... // comments
... unquoted: 'and you can quote me on that',
... singleQuotes: 'I can use "double quotes" here',
... lineBreaks: "Look, Mom! \
... No \\n's!",
... hexadecimal: 0xdecaf,
... leadingDecimalPoint: .8675309, andTrailing: 8675309.,
... positiveSign: +1,
... trailingComma: 'in objects', andIn: ['arrays',],
... "backwardsCompatible": "with JSON",
... }
... """
>>> chompjs.parse_js_object(data)
{'unquoted': 'and you can quote me on that', 'singleQuotes': 'I can use "double quotes" here', 'lineBreaks': "Look, Mom! No \n's!", 'hexadecimal': 912559, 'leadingDecimalPoint': 0.8675309, 'andTrailing': 8675309.0, 'positiveSign': '+1', 'trailingComma': 'in objects', 'andIn': ['arrays'], 'backwardsCompatible': 'with JSON'}If the input string is not yet escaped and contains a lot of \\ characters, then unicode_escape=True argument might help to sanitize it:
>>> chompjs.parse_js_object('{\\\"a\\\": 12}', unicode_escape=True)
{'a': 12}By default chompjs tries to start with first { or [ character it founds, omitting the rest:
>>> chompjs.parse_js_object('<div>...</div><script>foo = [1, 2, 3];</script><div>...</div>')
[1, 2, 3]Post-processed input is parsed using json.loads by default. A different loader such as orsjon can be used with loader argument:
>>> import orjson
>>> import chompjs
>>>
>>> chompjs.parse_js_object("{'a': 12}", loader=orjson.loads)
{'a': 12}loader_args and loader_kwargs arguments can be used to pass options to underlying loader function. For example for default json.loads you can pass down options such as strict or object_hook:
>>> import decimal
>>> import chompjs
>>> chompjs.parse_js_object('[23.2]', loader_kwargs={'parse_float': decimal.Decimal})
[Decimal('23.2')]In web scraping data often is not present directly inside HTML, but instead provided as an embedded JavaScript object that is later used to initialize the page, for example:
<html>
<head>...</head>
<body>
...
<script type="text/javascript">window.__PRELOADED_STATE__={"foo": "bar"}</script>
...
</body>
</html>Standard library function json.loads is usually sufficient to extract this data:
>>> # scrapy shell file:///tmp/test.html
>>> import json
>>> script_text = response.css('script:contains(__PRELOADED_STATE__)::text').re_first('__PRELOADED_STATE__=(.*)')
>>> json.loads(script_text)
{u'foo': u'bar'}The problem is that not all valid JavaScript objects are also valid JSONs. For example all those strings are valid JavaScript objects but not valid JSONs:
"{'a': 'b'}"is not a valid JSON because it uses'character to quote'{a: "b"}'is not a valid JSON because property name is not quoted at all'{"a": [1, 2, 3,]}'is not a valid JSON because there is an extra,character at the end of the array'{"a": .99}'is not a valid JSON because float value lacks a leading 0
As a result, json.loads fail to extract any of those:
>>> json.loads("{'a': 'b'}")
Traceback (most recent call last):
...
ValueError: Expecting property name: line 1 column 2 (char 1)
>>> json.loads('{a: "b"}')
Traceback (most recent call last):
...
ValueError: Expecting property name: line 1 column 2 (char 1)
>>> json.loads('{"a": [1, 2, 3,]}')
Traceback (most recent call last):
...
ValueError: No JSON object could be decoded
>>> json.loads('{"a": .99}')
Traceback (most recent call last):
...
json.decoder.JSONDecodeError: Expecting value: line 1 column 7 (char 6)chompjs library was designed to bypass this limitation, and it allows to scrape such JavaScript objects into proper Python dictionaries:
>>> import chompjs
>>>
>>> chompjs.parse_js_object("{'a': 'b'}")
{'a': 'b'}
>>> chompjs.parse_js_object('{a: "b"}')
{'a': 'b'}
>>> chompjs.parse_js_object('{"a": [1, 2, 3,]}')
{'a': [1, 2, 3]}
>>> chompjs.parse_js_object('{"a": .99}')
{'a': 0.99}Internally chompjs use a parser written in C to iterate over raw string, fixing its issues along the way. The final result is then passed down to standard library's json.loads, ensuring a high speed as compared to full-blown JavaScript parsers such as demjson.
>>> import json
>>> import _chompjs
>>>
>>> _chompjs.parse('{a: 1}')
'{"a":1}'
>>> json.loads(_)
{'a': 1}Pull requests are welcome.
To run unittests
$ tox
