Tools to Work with the 'Splash' JavaScript Rendering Service in R
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
boB Rudis 7263e6830c
tidy_har cleanup
4 years ago
R tidy_har cleanup 4 years ago
README_files/figure-markdown_github Added integration with HARtools to the TODO 7 years ago
img on the road to 7.0 4 years ago
man on the road to 7.0 4 years ago
tests forgot 8050 in docker 5 years ago
vignettes updated vignette 5 years ago
.Rbuildignore README 5 years ago
.gitignore updated vignette 5 years ago
.travis.yml master cran 5 years ago
CONDUCT.md code of conduct 7 years ago
CRAN-RELEASE README 5 years ago
DESCRIPTION on the road to 7.0 4 years ago
LICENSE license 5 years ago
LICENSE.md license 5 years ago
NAMESPACE on the road to 7.0 4 years ago
NEWS.md on the road to 7.0 4 years ago
README.Rmd tidy_har cleanup 4 years ago
README.md on the road to 7.0 4 years ago
codecov.yml appveyor tests 7 years ago
cran-comments.md updated vignette 5 years ago
splashr.Rproj initial commit 7 years ago

README.md

Travis-CI BuildStatus CoverageStatus CRAN\_Status\_Badge

splashr : Tools to Work with the ‘Splash’ JavaScript Rendering Service

TL;DR: This package works with Splash rendering servers which are really just a REST API & lua scripting interface to a QT browser. It’s an alternative to the Selenium ecosystem which was really engineered for application testing & validation.

Sometimes, all you need is a page scrape after javascript has been allowed to roam wild and free over meticulously crafted HTML tags. So, this package does not do everything Selenium can in pure R (though, the Lua interface is equally as powerful and accessible via R), but if you’re just trying to get a page back that needs javascript rendering, this is a nice, lightweight, consistent alternative.

It’s also an alternative to the somewhat abandoned phantomjs (which you can use in R within or without a Selenium context as it’s it’s own webdriver) and it may be useful to compare renderings between this package & phantomjs.

The package uses the stevedore package to orchestrate Docker on your system (if you have Docker and more on how to use the stevedore integration below) but you can also do get it running in Docker on the command-line with two commands:

sudo docker pull scrapinghub/splash:latest --disable-browser-caches
sudo docker run -p 5023:5023 -p 8050:8050 -p 8051:8051 scrapinghub/splash:latest --disable-browser-caches

Do whatever you Windows ppl do with Docker on your systems to make ^^ work.

Folks super-new to Docker on Unix-ish platforms should make sure to do:

sudo groupadd docker
sudo usermod -aG docker $USER

($USER is your username and shld be defined for you in the environment)

If using the stevedore package you can use the convience wrappers in this pacakge:

install_splash()
splash_container <- start_splash()

and then run:

stop_splash(splash_container)

when done. All of that happens on your localhost and you will not need to specify splash_obj to many of the splashr functions if you’re running Splash in this default configuration as long as you use named parameters. You can also use the pre-defined splash_local object if you want to use positional parameters.

Now, you can run Selenium in Docker, so this is not unique to Splash. But, a Docker context makes it so that you don’t have to run or maintain icky Python stuff directly on your system. Leave it in the abandoned warehouse district where it belongs.

All you need for this package to work is a running Splash instance. You provide the host/port for it and it’s scrape-tastic fun from there!

About Splash

‘Splash’ https://github.com/scrapinghub/splash is a javascript rendering service. It’s a lightweight web browser with an ‘HTTP’ API, implemented in Python using ‘Twisted’and ’QT’ [and provides some of the core functionality of the ‘RSelenium’ or ‘seleniumPipes’ R packages but with a Java-free footprint]. The (twisted) ‘QT’ reactor is used to make the sever fully asynchronous allowing to take advantage of ‘webkit’ concurrency via QT main loop. Some of Splash features include the ability to process multiple webpages in parallel; retrieving HTML results and/or take screenshots; disabling images or use Adblock Plus rules to make rendering faster; executing custom JavaScript in page context; getting detailed rendering info in HAR format.

The following functions are implemented:

  • render_html: Return the HTML of the javascript-rendered page.
  • render_har: Return information about Splash interaction with a website in HAR format.
  • render_jpeg: Return a image (in JPEG format) of the javascript-rendered page.
  • render_png: Return a image (in PNG format) of the javascript-rendered page.
  • execute_lua: Execute a custom rendering script and return a result.
  • splash: Configure parameters for connecting to a Splash server
  • install_splash: Retrieve the Docker image for Splash
  • start_splash: Start a Splash server Docker container
  • stop_splash: Stop a running a Splash server Docker container

Mini-DSL (domain-specific language). These can be used to create a “script” without actually scripting in Lua. They are a less-powerful/configurable set of calls than what you can make with a full Lua function but the idea is to have it take care of very common but simple use-cases, like waiting a period of time before capturing a HAR/HTML/PNG image of a site:

  • splash_plugins: Enable or disable browser plugins (e.g. Flash).
  • splash_click: Trigger mouse click event in web page.
  • splash_focus: Focus on a document element provided by a CSS selector
  • splash_images: Enable/disable images
  • splash_response_body: Enable or disable response content tracking.
  • splash_go: Go to an URL.
  • splash_wait: Wait for a period time
  • splash_har: Return information about Splash interaction with a website in HAR format.
  • splash_html: Return a HTML snapshot of a current page.
  • splash_png: Return a screenshot of a current page in PNG format.
  • splash_press: Trigger mouse press event in web page.
  • splash_release: Trigger mouse release event in web page.
  • splash_send_keys: Send keyboard events to page context.
  • splash_send_text: Send text as input to page context, literally, character by character.
  • splash_user_agent: Overwrite the User-Agent header for all further requests. NOTE: There are many “helper” user agent strings to go with splash_user_agent. Look for objects in splashr starting with ua_.

httr helpers. These help turn various bits of splashr objects into httr-ish things:

  • as_req: Turn a HAR response entry into a working httr function you can use to make a request with
  • as_request: Turn a HAR response entry into an httr response-like object (i.e. you can use httr::content() on it)

Helpers:

  • tidy_har: Turn a gnHARly HAR object into a tidy data frame
  • get_body_size: Retrieve size of content | body | headers
  • get_content_sie: Retrieve size of content | body | headers
  • get_content_type Retrieve or test content type of a HAR request object
  • get_headers_size Retrieve size of content | body | headers
  • is_binary: Retrieve or test content type of a HAR request object
  • is_content_type: Retrieve or test content type of a HAR request object
  • is_css: Retrieve or test content type of a HAR request object
  • is_gif: Retrieve or test content type of a HAR request object
  • is_html: Retrieve or test content type of a HAR request object
  • is_javascript: Retrieve or test content type of a HAR request object
  • is_jpeg: Retrieve or test content type of a HAR request object
  • is_json: Retrieve or test content type of a HAR request object
  • is_plain: Retrieve or test content type of a HAR request object
  • is_png: Retrieve or test content type of a HAR request object
  • is_svg: Retrieve or test content type of a HAR request object
  • is_xhr: Retrieve or test content type of a HAR request object
  • is_xml: Retrieve or test content type of a HAR request object

Some functions from HARtools are imported/exported and %>% is imported/exported.

TODO

Suggest more in a feature req!

  • Implement render.json
  • Implement “file rendering”
  • Implement execute (you can script Splash!)
  • Add integration with HARtools
  • Possibly writing R function wrappers to install/start/stop Splash which would also support enabling javascript profiles, request filters and proxy profiles from with R directly, using harbor
  • Re-implement render_file()
  • Testing results with all combinations of parameters

Installation

# CRAN
install.packages("splashr")

# DEV
# See DESCRIPTION for non-CINC-provided dependencies 
install.packages("splashr", repos = c("https://cinc.rud.is/"))

Usage

NOTE: ALL of these examples assume Splash is running in the default configuraiton on localhost (i.e. started with start_splash() or the docker example commands) unless otherwise noted.

library(splashr)
library(magick)
library(rvest)
library(anytime)
library(tidyverse)

# current verison
packageVersion("splashr")
## [1] '0.7.0'
splash_active()
## [1] TRUE
splash_debug()
## List of 7
##  $ active  : list()
##  $ argcache: int 0
##  $ fds     : int 59
##  $ leaks   :List of 5
##   ..$ Deferred   : int 20
##   ..$ DelayedCall: int 3
##   ..$ LuaRuntime : int 1
##   ..$ QTimer     : int 1
##   ..$ Request    : int 1
##  $ maxrss  : int 219096
##  $ qsize   : int 0
##  $ url     : chr "http://localhost:8050"
##  - attr(*, "class")= chr [1:2] "splash_debug" "list"
## NULL

Notice the difference between a rendered HTML scrape and a non-rendered one:

render_html(url = "http://marvel.com/universe/Captain_America_(Steve_Rogers)")
## {html_document}
## <html lang="en">
## [1] <head class="at-element-marker">\n<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">\n<meta char ...
## [2] <body class="terrigen">\n<div id="__next"><div id="terrigen-page" class="page"><div id="page-wrapper" class="page ...
xml2::read_html("http://marvel.com/universe/Captain_America_(Steve_Rogers)")
## {html_document}
## <html lang="en">
## [1] <head>\n<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">\n<meta charset="utf-8" class="next-he ...
## [2] <body class="terrigen">\n<div id="__next"><div id="terrigen-page" class="page"><div id="page-wrapper" class="page ...

You can also profile pages:

render_har(url = "http://www.poynter.org/") -> har

print(har)
## --------HAR VERSION-------- 
## HAR specification version: 1.2 
## --------HAR CREATOR-------- 
## Created by: Splash 
## version: 3.4.1 
## --------HAR BROWSER-------- 
## Browser: QWebKit 
## version: 602.1 
## --------HAR PAGES-------- 
## Page id: 1 , Page title: Poynter - Poynter 
## --------HAR ENTRIES-------- 
## Number of entries: 270 
## REQUESTS: 
## Page: 1 
## Number of entries: 270 
##   -  http://www.poynter.org/ 
##   -  https://www.poynter.org/ 
##   -  https://www.googletagservices.com/tag/js/gpt.js 
##   -  https://www.poynter.org/wp-includes/js/wp-emoji-release.min.js?ver=5.3.2 
##   -  https://www.poynter.org/wp-includes/css/dist/block-library/style.min.css?ver=5.3.2 
##      ........ 
##   -  https://www.poynter.org/wp-content/themes/elumine-child/assets/fonts/PoynterOSDisplay/PoynterOSDisplay-BoldItalic... 
##   -  https://securepubads.g.doubleclick.net/gampad/ads?gdfp_req=1&pvsid=3405643238502098&correlator=142496486392385&ou... 
##   -  https://securepubads.g.doubleclick.net/gpt/pubads_impl_rendering_2020021101.js 
##   -  https://tpc.googlesyndication.com/safeframe/1-0-37/html/container.html 
##   -  https://static.chartbeat.com/js/chartbeat.js
tidy_har(har)
## # A tibble: 270 x 20
##    status started total_time page_ref timings req_url req_method req_http_version req_hdr_size req_headers req_cookies
##     <int> <chr>        <int> <chr>    <I<lis> <chr>   <chr>      <chr>                   <int> <I<list>>   <I<list>>  
##  1    301 2020-0…        171 1        <tibbl… http:/… GET        HTTP/1.1                  189 <tibble [2… <tibble [0…
##  2    200 2020-0…        238 1        <tibbl… https:… GET        HTTP/1.1                  189 <tibble [2… <tibble [0…
##  3    200 2020-0…        135 1        <tibbl… https:… GET        HTTP/1.1                  164 <tibble [3… <tibble [0…
##  4    200 2020-0…         93 1        <tibbl… https:… GET        HTTP/1.1                  164 <tibble [3… <tibble [0…
##  5    200 2020-0…        158 1        <tibbl… https:… GET        HTTP/1.1                  179 <tibble [3… <tibble [0…
##  6    200 2020-0…        218 1        <tibbl… https:… GET        HTTP/1.1                  179 <tibble [3… <tibble [0…
##  7    200 2020-0…        262 1        <tibbl… https:… GET        HTTP/1.1                  179 <tibble [3… <tibble [0…
##  8    200 2020-0…        265 1        <tibbl… https:… GET        HTTP/1.1                  179 <tibble [3… <tibble [0…
##  9    200 2020-0…        266 1        <tibbl… https:… GET        HTTP/1.1                  179 <tibble [3… <tibble [0…
## 10    200 2020-0…        269 1        <tibbl… https:… GET        HTTP/1.1                  179 <tibble [3… <tibble [0…
## # … with 260 more rows, and 9 more variables: resp_url <chr>, resp_rdrurl <chr>, resp_type <chr>, resp_size <int>,
## #   resp_cookies <I<list>>, resp_headers <I<list>>, resp_encoding <chr>, resp_content_size <dbl>, resp_content <chr>

You can use HARtools::HARviewer — which this pkg import/exports — to get view the HAR in an interactive HTML widget.

Full web page snapshots are easy-peasy too:

render_png(url = "http://www.marveldirectory.com/individuals/c/captainamerica.htm")

render_jpeg(url = "http://static2.comicvine.com/uploads/scale_small/3/31666/5052983-capasr2016001-eptingvar-18bdb.jpg") 

Executing custom Lua scripts

lua_ex <- '
function main(splash)
  splash:go("http://rud.is/b")
  splash:wait(0.5)
  local title = splash:evaljs("document.title")
  return {title=title}
end
'

splash_local %>% execute_lua(lua_ex) -> res

rawToChar(res) %>% 
  jsonlite::fromJSON()
## $title
## [1] "rud.is | \"In God we trust. All others must bring data\""

Interacting With Flash sites

splash_local %>%
  splash_plugins(TRUE) %>%
  splash_go("https://gis.cdc.gov/GRASP/Fluview/FluHospRates.html") %>%
  splash_wait(4) %>%
  splash_click(460, 550) %>%
  splash_wait(2) %>%
  splash_click(230, 85) %>%
  splash_wait(2) %>%
  splash_png()
stop_splash(splash_vm)

Code of Conduct

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.