A Python Framework For Using HAR Files To Analyze Web Pages.
The haralyzer module contains two classes for analyzing web pages based
on a HAR file. HarParser()
represents a full file (which might have
multiple pages), and HarPage()
represents a single page from said file.
HarParser
has a couple of helpful methods for analyzing single entries
from a HAR file, but most of the pertinent functions are inside of the page
object.
haralyzer
was designed to be easy to use, but you can also access more
powerful functions directly.
The HarParser
takes a single argument of a dict
representing the JSON
of a full HAR file. It has the same properties of the HAR file, EXCEPT that each
page in HarParser.pages is a HarPage object:
import json from haralyzer import HarParser, HarPage with open('har_data.har', 'r') as f: har_parser = HarParser(json.loads(f.read())) print har_parser.browser # {u'name': u'Firefox', u'version': u'25.0.1'} print har_parser.hostname # 'humanssuck.net' for page in har_parser.pages: assert isinstance(page, HarPage, None) # returns True for each
The HarPage
object contains most of the goods you need to easily analyze a
page. It has helper methods that are accessible, but most of the data you need is
in properties for easy access. You can create a HarPage object directly by giving
it the page ID (yes, I know it is stupid, it's just how HAR is organized), and either
a HarParser
with har_parser=parser, or a dict
representing the JSON of a full HAR
file (see example above) with har_data=har_data:
import json From haralyzer import HarPage with open('har_data.har', 'r') as f: har_page = HarPage('page_3', har_data=json.loads(f.read())) ### GET BASIC INFO har_page.hostname # 'humanssuck.net' har_page.url $ 'http://humanssuck.net/about/' ### WORK WITH LOAD TIMES (all load times are in ms) ### # Get image load time in milliseconds as rendered by the browser har_page.image_load_time # prints 713 # We could do this with 'css', 'js', 'html', 'audio', or 'video' ### WORK WITH SIZES (all sizes are in bytes) ### # Get the total page size (with all assets) har_page.page_size # prints 2423765 # Get the total image size har_page.image_size # prints 733488 # We could do this with 'css', 'js', 'html', 'audio', or 'video' # Get the transferred sizes (works only with HAR files, generated with Chrome) har_page.page_size_trans har_page.image_size_trans har_page.css_size_trans har_page.text_size_trans har_page.js_size_trans har_page.audio_size_trans har_page.video_size_trans
IMPORTANT NOTE - Technically, the page_id attribute of a single entry in a HAR file is optional. As such, if your HAR file contains entries that do not map to a page, an additional page will be created with an ID of unknown. This "fake page" will contain all such entries. Since it is not a real page, it does not have attributes for things like time to first byte or page load, and will return None.
The MutliHarParser
takes a list
of dict
, each of which represents the JSON
of a full HAR file. The concept here is that you can provide multiple HAR files of the
same page (representing multiple test runs) and the MultiHarParser
will provide
aggregate results for load times:
import json from haralyzer import HarParser, HarPage test_runs = [] with open('har_data1.har', 'r') as f1: test_runs.append( (json.loads( f1.read() ) ) with open('har_data2.har', 'r') as f2: test_runs.append( (json.loads( f2.read() ) ) multi_har_parser = MultiHarParser(har_data=test_runs) # Get the mean for the time to first byte of all runs in MS print multi_har_parser.time_to_first_byte # 70 # Get the total page load time mean for all runs in MS print multi_har_parser.load_time # 150 # Get the javascript load time mean for all runs in MS print multi_har_parser.js_load_time # 50 # You can get the standard deviation for any of these as well # Let's get the standard deviation for javascript load time print multi_har_parser.get_stdev('js') # 5 # We can also do that with 'page' or 'ttfb' (time to first byte) print multi_har_parser.get_stdev('page') # 11 print multi_har_parser.get_stdev('ttfb') # 10 ### DECIMAL PRECISION ### # You will notice that all of the results are above. That is because # the default decimal precision for the multi parser is 0. However, you # can pass whatever you want into the constructor to control this. multi_har_parser = MultiHarParser(har_data=test_runs, decimal_precision=2) print multi_har_parser.time_to_first_byte # 70.15
HarPage
includes a lot of helpful properties, but they are all
easily produced using the public methods of HarParser
and HarPage
:
import json from haralyzer import HarPage with open('har_data.har', 'r') as f: har_page = HarPage('page_3', har_data=json.loads(f.read())) ### ACCESSING FILES ### # You can get a JSON representation of all assets using HarPage.entries # for entry in har_page.entries: if entry['startedDateTime'] == 'whatever I expect': ... do stuff ... # It also has methods for filtering assets # # Get a collection of entries that were images in the 2XX status code range # entries = har_page.filter_entries(content_type='image.*', status_code='2.*') # This method can filter by: # * content_type ('application/json' for example) # * status_code ('200' for example) # * request_type ('GET' for example) # * http_version ('HTTP/1.1' for example) # It will use a regex by default, but you can also force a literal string match by passing regex=False # Get the size of the collection we just made # collection_size = har_page.get_total_size(entries) # We can also access files by type with a property # for js_file in har_page.js_files: ... do stuff .... ### GETTING LOAD TIMES ### # Get the BROWSER load time for all images in the 2XX status code range # load_time = har_page.get_load_time(content_type='image.*', status_code='2.*') # Get the TOTAL load time for all images in the 2XX status code range # load_time = har_page.get_load_time(content_type='image.*', status_code='2.*', asynchronous=False)
This could potentially be out of date, so please check out the sphinx docs.
All of the HarPage methods above leverage stuff from the HarParser,
some of which can be useful for more complex operations. They either
operate on a single entry (from a HarPage) or a list
of entries:
import json from haralyzer import HarParser with open('har_data.har', 'r') as f: har_parser = HarParser(json.loads(f.read())) for page in har_parser.pages: for entry in page.entries: ### MATCH HEADERS ### if har_parser.match_headers(entry, 'Content-Type', 'image.*'): print 'This would appear to be an image' ### MATCH REQUEST TYPE ### if har_parser.match_request_type(entry, 'GET'): print 'This is a GET request' ### MATCH STATUS CODE ### if har_parser.match_status_code(entry, '2.*'): print 'Looks like all is well in the world'
The last helper function of HarParser
requires it's own section, because it
is odd, but can be helpful, especially for creating charts and reports.
It can create an asset timeline, which gives you back a dict
where each
key is a datetime
object, and the value is a list
of assets that were
loading at that time. Each value of the list
is a dict
representing
an entry from a page.
It takes a list
of entries to analyze, so it assumes that you have
already filtered the entries you want to know about:
import json from haralyzer import HarParser with open('har_data.har', 'r') as f: har_parser = HarParser(json.loads(f.read())) ### CREATE A TIMELINE OF ALL THE ENTRIES ### entries = [] for page in har_parser.pages: for entry in page.entries: entries.append(entry) timeline = har_parser.create_asset_timeline(entries) for key, value in timeline.items(): print(type(key)) # <type 'datetime.datetime'> print(key) # 2015-02-21 19:15:41.450000-08:00 print(type(value)) # <type 'list'> print(value) # Each entry in the list is an asset from the page # [{u'serverIPAddress': u'157.166.249.67', u'cache': {}, u'startedDateTime': u'2015-02-21T19:15:40.351-08:00', u'pageref': u'page_3', u'request': {u'cookies':............................
With this, you can examine the timeline for any number of assets. Since the key is a datetime
object, this is a heavy operation. We could always change this in the future, but for now,
limit the assets you give this method to only what you need to examine.