This fork provides a quick fix for pagination.py. In the current official release (2.4.0) pagination doesn't return all results pages. This fork provides a quick fix for this by changing the stop condition for the pagination iterator.
This Python package is meant to scrape and parse search resutls from Google, Bing, Baidu, Yandex, Yahoo, Home depot, Ebay and more.. using SerpApi.
The following services are provided:
- Search API
- Search Archive API
- Account API
- Location API (Google Only)
SerpApi provides a script builder to get you started quickly.
Python 3.7+
pip install google-search-resultsLink to the python package page
from serpapi import GoogleSearch
search = GoogleSearch({
"q": "coffee",
"location": "Austin,Texas",
"api_key": "<your secret api key>"
})
result = search.get_dict()This example runs a search about "coffee" using your secret api key.
The SerpApi service (backend)
- searches on Google using the search: q = "coffee"
- parses the messy HTML responses
- return a standardizes JSON response The GoogleSearch class
- Format the request
- Execute GET http request against SerpApi service
- Parse JSON response into a dictionary Et voila..
Alternatively, you can search:
- Bing using BingSearch class
- Baidu using BaiduSearch class
- Yahoo using YahooSearch class
- Ebay using EbaySearch class
- Yandex using YandexSearch class
- HomeDepot using HomeDepotSearch class
- GoogleScholar using GoogleScholarSearch class
See the playground to generate your code.
- Google Search Results in Python
- Installation
- Quick start
- Summary
- Google Search API capability
- How to set SERP API key
- Example by specification
- Location API
- Search Archive API
- Account API
- Search Bing
- Search Baidu
- Search Yandex
- Search Yahoo
- Search Ebay
- Search Home depot
- Search Youtube
- Search Google Scholar
- Generic search with SerpApiClient
- Search Google Images
- Search Google News
- Search Google Shopping
- Google Search By Location
- Batch Asynchronous Searches
- Python object as a result
- Python paginate using iterator
- Error management
- Change log
- Conclusion
Source code.
params = {
"q": "coffee",
"location": "Location Requested",
"device": "desktop|mobile|tablet",
"hl": "Google UI Language",
"gl": "Google Country",
"safe": "Safe Search Flag",
"num": "Number of Results",
"start": "Pagination Offset",
"api_key": "Your SERP API Key",
# To be match
"tbm": "nws|isch|shop",
# To be search
"tbs": "custom to be search criteria",
# allow async request
"async": "true|false",
# output format
"output": "json|html"
}
# define the search search
search = GoogleSearch(params)
# override an existing parameter
search.params_dict["location"] = "Portland"
# search format return as raw html
html_results = search.get_html()
# parse results
# as python Dictionary
dict_results = search.get_dict()
# as JSON using json package
json_results = search.get_json()
# as dynamic Python object
object_result = search.get_object()Link to the full documentation
see below for more hands on examples.
You can get an API key here if you don't already have one: https://serpapi.com/users/sign_up
The SerpApi api_key can be set globally:
GoogleSearch.SERP_API_KEY = "Your Private Key"The SerpApi api_key can be provided for each search:
query = GoogleSearch({"q": "coffee", "serp_api_key": "Your Private Key"})We love true open source, continuous integration and Test Drive Development (TDD). We are using RSpec to test our infrastructure around the clock to achieve the best QoS (Quality Of Service).
The directory test/ includes specification/examples.
Set your api key.
export API_KEY="your secret key"Run test
make testfrom serpapi import GoogleSearch
search = GoogleSearch({})
location_list = search.get_location("Austin", 3)
print(location_list)it prints the first 3 location matching Austin (Texas, Texas, Rochester)
[ { 'canonical_name': 'Austin,TX,Texas,United States',
'country_code': 'US',
'google_id': 200635,
'google_parent_id': 21176,
'gps': [-97.7430608, 30.267153],
'id': '585069bdee19ad271e9bc072',
'keys': ['austin', 'tx', 'texas', 'united', 'states'],
'name': 'Austin, TX',
'reach': 5560000,
'target_type': 'DMA Region'},
...]The search result are stored in temporary cached. The previous search can be retrieve from the the cache for free.
from serpapi import GoogleSearch
search = GoogleSearch({"q": "Coffee", "location": "Austin,Texas"})
search_result = search.get_dictionary()
assert search_result.get("error") == None
search_id = search_result.get("search_metadata").get("id")
print(search_id)Now let retrieve the previous search from the archive.
archived_search_result = GoogleSearch({}).get_search_archive(search_id, 'json')
print(archived_search_result.get("search_metadata").get("id"))it prints the search result from the archive.
from serpapi import GoogleSearch
search = GoogleSearch({})
account = search.get_account()it prints your account information.
from serpapi import BingSearch
search = BingSearch({"q": "Coffee", "location": "Austin,Texas"})
data = search.get_dict()this code prints baidu search results for coffee as a Dictionary.
https://serpapi.com/bing-search-api
from serpapi import BaiduSearch
search = BaiduSearch({"q": "Coffee"})
data = search.get_dict()this code prints baidu search results for coffee as a Dictionary. https://serpapi.com/baidu-search-api
from serpapi import YandexSearch
search = YandexSearch({"text": "Coffee"})
data = search.get_dict()this code prints yandex search results for coffee as a Dictionary.
https://serpapi.com/yandex-search-api
from serpapi import YahooSearch
search = YahooSearch({"p": "Coffee"})
data = search.get_dict()this code prints yahoo search results for coffee as a Dictionary.
https://serpapi.com/yahoo-search-api
from serpapi import EbaySearch
search = EbaySearch({"_nkw": "Coffee"})
data = search.get_dict()this code prints ebay search results for coffee as a Dictionary.
https://serpapi.com/ebay-search-api
from serpapi import HomeDepotSearch
search = HomeDepotSearch({"q": "chair"})
data = search.get_dict()this code prints home depot search results for chair as Dictionary.
https://serpapi.com/home-depot-search-api
from serpapi import HomeDepotSearch
search = YoutubeSearch({"q": "chair"})
data = search.get_dict()this code prints youtube search results for chair as Dictionary.
https://serpapi.com/youtube-search-api
from serpapi import GoogleScholarSearch
search = GoogleScholarSearch({"q": "Coffee"})
data = search.get_dict()this code prints Google Scholar search results.
from serpapi import SerpApiClient
query = {"q": "Coffee", "location": "Austin,Texas", "engine": "google"}
search = SerpApiClient(query)
data = search.get_dict()This class enables to interact with any search engine supported by SerpApi.com
from serpapi import GoogleSearch
search = GoogleSearch({"q": "coffe", "tbm": "isch"})
for image_result in search.get_dict()['images_results']:
link = image_result["original"]
try:
print("link: " + link)
# wget.download(link, '.')
except:
passthis code prints all the images links, and download image if you un-comment the line with wget (linux/osx tool to download image).
This tutorial covers more ground on this topic. https://github.com/serpapi/showcase-serpapi-tensorflow-keras-image-training
from serpapi import GoogleSearch
search = GoogleSearch({
"q": "coffe", # search search
"tbm": "nws", # news
"tbs": "qdr:d", # last 24h
"num": 10
})
for offset in [0,1,2]:
search.params_dict["start"] = offset * 10
data = search.get_dict()
for news_result in data['news_results']:
print(str(news_result['position'] + offset * 10) + " - " + news_result['title'])this script prints the first 3 pages of the news title for the last 24h.
from serpapi import GoogleSearch
search = GoogleSearch({
"q": "coffe", # search search
"tbm": "shop", # news
"tbs": "p_ord:rv", # last 24h
"num": 100
})
data = search.get_dict()
for shopping_result in data['shopping_results']:
print(shopping_result['position']) + " - " + shopping_result['title'])this script prints all the shopping results order by review order.
With SerpApi, we can build Google search from anywhere in the world. This code is looking for the best coffee shop per city.
from serpapi import GoogleSearch
for city in ["new york", "paris", "berlin"]:
location = GoogleSearch({}).get_location(city, 1)[0]["canonical_name"]
search = GoogleSearch({
"q": "best coffee shop", # search search
"location": location,
"num": 1,
"start": 0
})
data = search.get_dict()
top_result = data["organic_results"][0]["title"]We do offer two ways to boost your searches thanks to async parameter.
- Blocking - async=false - it's more compute intensive because the search would need to hold many connections. (default)
- Non-blocking - async=true - it's way to go for large amount of query submitted by batch (recommended)
# Operating system
import os
# regular expression library
import re
# safe queue (named Queue in python2)
from queue import Queue
# Time utility
import time
# SerpApi search
from serpapi import GoogleSearch
# store searches
search_queue = Queue()
# SerpApi search
search = GoogleSearch({
"location": "Austin,Texas",
"async": True,
"api_key": os.getenv("API_KEY")
})
# loop through a list of companies
for company in ['amd', 'nvidia', 'intel']:
print("execute async search: q = " + company)
search.params_dict["q"] = company
result = search.get_dict()
if "error" in result:
print("oops error: ", result["error"])
continue
print("add search to the queue where id: ", result['search_metadata'])
# add search to the search_queue
search_queue.put(result)
print("wait until all search statuses are cached or success")
# Create regular search
while not search_queue.empty():
result = search_queue.get()
search_id = result['search_metadata']['id']
# retrieve search from the archive - blocker
print(search_id + ": get search from archive")
search_archived = search.get_search_archive(search_id)
print(search_id + ": status = " +
search_archived['search_metadata']['status'])
# check status
if re.search('Cached|Success',
search_archived['search_metadata']['status']):
print(search_id + ": search done with q = " +
search_archived['search_parameters']['q'])
else:
# requeue search_queue
print(search_id + ": requeue search")
search_queue.put(result)
# wait 1s
time.sleep(1)
print('all searches completed')This code shows how to run searches asynchronously. The search parameters must have {async: True}. This indicates that the client shouldn't wait for the search to be completed. The current thread that executes the search is now non-blocking which allows to execute thousand of searches in seconds. The SerpApi backend will do the processing work. The actual search result is defer to a later call from the search archive using get_search_archive(search_id). In this example the non-blocking searches are persisted in a queue: search_queue. A loop through the search_queue allows to fetch individual search result. This process can be easily multithreaded to allow a large number of concurrent search requests. To keep thing simple, this example does only explore search result one at a time (single threaded).
The search results can be automatically wrapped in dynamically generated Python object. This solution offers a more dynamic solution fully Oriented Object Programming approach over the regular Dictionary / JSON data structure.
from serpapi import GoogleSearch
search = GoogleSearch({"q": "Coffee", "location": "Austin,Texas"})
r = search.get_object()
assert type(r.organic_results), list
assert r.organic_results[0].title
assert r.search_metadata.id
assert r.search_metadata.google_url
assert r.search_parameters.q, "Coffee"
assert r.search_parameters.engine, "google"Let's collect links accross multiple search result pages.
# to get 2 pages
start = 0
end = 40
page_size = 10
# basic search parameters
parameter = {
"q": "coca cola",
"tbm": "nws",
"api_key": os.getenv("API_KEY"),
# optional pagination parameter
# the pagination method can take argument directly
"start": start,
"end": end,
"num": page_size
}
# as proof of concept
# urls collects
urls = []
# initialize a search
search = GoogleSearch(parameter)
# create a python generator using parameter
pages = search.pagination()
# or set custom parameter
pages = search.pagination(start, end, page_size)
# fetch one search result per iteration
# using a basic python for loop
# which invokes python iterator under the hood.
for page in pages:
print(f"Current page: {page['serpapi_pagination']['current']}")
for news_result in page["news_results"]:
print(f"Title: {news_result['title']}\nLink: {news_result['link']}\n")
urls.append(news_result['link'])
# check if the total number pages is as expected
# note: the exact number if variable depending on the search engine backend
if len(urls) == (end - start):
print("all search results count match!")
if len(urls) == len(set(urls)):
print("all search results are unique!")Examples to fetch links with pagination: test file, online IDE
SerpAPI keeps error mangement very basic.
- backend service error or search fail
- client error
If it's a backend error, a simple message error is returned as string in the server response.
from serpapi import GoogleSearch
search = GoogleSearch({"q": "Coffee", "location": "Austin,Texas", "api_key": "<secret_key>"})
data = search.get_json()
assert data["error"] == NoneIn some case, there is more details availabel in the data object.
If it's client error, then a SerpApiClientException is raised.
2021-09-01 @ 2.4.1
- raise SerpApiClientException instead of raw string in order to follow Python guideline 3.5+
- add more unit error tests for serp_api_client 2021-07-26 @ 2.4.0
- add page size support using num parameter
- add youtube search engine 2021-06-05 @ 2.3.0
- add pagination support 2021-04-28 @ 2.2.0
- add get_response method to provide raw requests.Response object 2021-04-04 @ 2.1.0
- Add home depot search engine
- get_object() returns dynamic Python object 2020-10-26 @ 2.0.0
- Reduce class name to Search
- Add get_raw_json 2020-06-30 @ 1.8.3
- simplify import
- improve package for python 3.5+
- add support for python 3.5 and 3.6 2020-03-25 @ 1.8
- add support for Yandex, Yahoo, Ebay
- clean-up test 2019-11-10 @ 1.7.1
- increase engine parameter priority over engine value set in the class 2019-09-12 @ 1.7
- Change namespace "from lib." instead: "from serpapi import GoogleSearch"
- Support for Bing and Baidu 2019-06-25 @ 1.6
- New search engine supported: Baidu and Bing
SerpApi supports all the major search engines. Google has the more advance support with all the major services available: Images, News, Shopping and more.. To enable a type of search, the field tbm (to be matched) must be set to:
- isch: Google Images API.
- nws: Google News API.
- shop: Google Shopping API.
- any other Google service should work out of the box.
- (no tbm parameter): regular Google search.
The field tbs allows to customize the search even more.