Skip to content

Latest commit

 

History

History
271 lines (194 loc) · 9.85 KB

README.md

File metadata and controls

271 lines (194 loc) · 9.85 KB
Raw

Snapsite

Visual regression test your whole website.

  • Snapshot your site for the first time with --go.
  • Come back later and --test.
  • Look at the --report.
  • Are the failed tests actually just improvements? Then --approve.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. Should you want to see a presentation have a look here

Prerequisites

  • Node
  • Google Chrome
  • Only tested on Windows so far, but should work wherever Chrome runs

Installation

In a directory with no spaces in the path, run:

git clone https://github.com/ie/snapsite-vrt.git
cd snapsite-vrt
npm install

And you're ready.

Dev Usage

After a git pull and build on your desired website, go through these steps:

  1. Run node snapsite.js --crawl examplesite.local
  2. Run node snapsite.js --reference examplesite.local
  3. Do some web development
  4. Run node snapsite.js --test examplesite.local
  5. Run node snapsite.js --report examplesite.local
  6. Handle reported failures:
    • If these failures are actually improvements, run node snapsite.js --approve examplesite.local
    • If these failures are real, go back to (3) and fix your broken tests.
  7. Once there are no failures, push your changes / submit a pull request.

CI Usage

  1. Install on a server with plenty of disk space
  2. Run node snapsite.js --crawl examplesitecom.au once
    • Optional: using your own script, filter out any undesirable URLs from crawled-urls.log
  3. Run node snapsite.js --reference examplesite.com.au once
  4. Nightly, run node snapsite.js --test examplesite.com.au (failed tests will yield a non-zero exit code)
  5. Handle reported differences:
    • Last report results are logged in sites/.../backstop_data/ci_report/xunit.xml; make this file accessible to devs so they can investigate.

Examples

Go crawl and capture reference screenshots for all of examplesite.com.au

node snapsite.js --go examplesite.com.au

Note: you can safely ^C in the middle of this.

Crawl examplesite.com.au

node snapsite.js --crawl examplesite.com.au

Note: you can safely ^C in the middle of this.

Reference image capture for all crawled URLs for examplesite.com.au

node snapsite.js --reference examplesite.com.au

Test image capture for all referenced URLs for examplesite.com.au

node snapsite.js --test examplesite.com.au

Present report of most recent test results for examplesite.com.au

node snapsite.js --report examplesite.com.au

Delete all data for examplesite.com.au

node snapsite.js --delete examplesite.com.au

Actions

--help (-h)

node snapsite.js or

node snapsite.js --help

Good to know it's there if you need it.

--go (-g)

node snapsite.js --go [-f] <domain> [-o <domain>|-O <relative-path>]

Crawls a whole website, while capturing reference images to be used as the basis for test success or failure. Essentially runs a --crawl, and every few URLs does a little --reference.

  • Use -f to delete the site directory before running
  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path
  • Outputs to ./sites/subdomain_domain_com_au/
  • HTML is logged to ./sites/.../html/.
  • Crawled URLs are logged to ./sites/.../crawled-urls.log.
  • Referenced URLs are logged to ./sites/.../referenced-urls.log.
  • Reference images are saved to ./sites/.../backstop_data/bitmaps_reference/

--crawl (-c)

node snapsite.js --crawl [-f] <domain> [-o <domain>|-O <relative-path>]

Crawls a whole website, starting with the URL you give it. Limits the crawl to just the domain of that URL and its subdomains.

  • Use -f to delete the site directory before running
  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path
  • Outputs to ./sites/subdomain_domain_com_au/
  • HTML is logged to ./sites/.../html/.
  • Crawled URLs are logged to ./sites/.../crawled-urls.log.

--reference (-r)

node snapsite.js --reference [-f] (<domain>|-u <url1> <url2> ...) [-o <domain>|-O <relative-path>]

Captures images to be used as the basis for test success or failure.

  • Use -u instead of a domain to reference specific URLs only
  • Use -f to delete reference files and test files before running
  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path
  • Outputs to ./sites/.../referenced-urls.log and ./sites/.../backstop_data/bitmaps_reference/

--test (-t)

node snapsite.js --test (<domain>|-u <url1> <url2> ...) [-o <domain>|-O <relative-path>]

Captures images again, and compares them to the reference images.

  • Use -u instead of a domain to test specific URLs only
  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path
  • Outputs to ./sites/.../tested-urls.log and ./sites/.../backstop_data/bitmaps_test/<date>-<time>/

--report (-p)

node snapsite.js --report <domain> [-o <domain>|-O <relative-path>]

Presents a report for you to inspect the results of the last test, eyeballing the nature of the failures to see if they are "real".

  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path

--approve (-a)

node snapsite.js --approve <domain> [-o <domain>|-O <relative-path>]

If all "failures" in the last test were intentional changes, then you should approve them. This copies any "failed but approved" test images over the top of existing reference images.

  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path

--delete (-d)

node snapsite.js --delete <domain> [-o <domain>|-O <relative-path>]

Deletes the site directory for the given domain.

  • Use -o to override the directory using a domain name
  • Use -O to override the directory using a relative path

--version (-v)

node snapsite.js --version

Known bugs

  • Some sites refuse to reference at all (e.g. lexus.com.au) due to a bug in the way we freeze VH units to pixels. We're working on it.

  • Sometimes, some URLs can be skipped during a reference/test. The cause is currently unknown.

  • Failed --go and --reference references are still added to referenced-urls.log. As a workaround you can use --reference -u url1 url2 ... -o examplesite.com.au to redo just those references.

Known limitations

  • --crawl logs to crawled-urls.log as it progresses, but --reference and test don't log to referenced-urls.log / tested-urls.log until they have completed.

  • SuperCrawler cannot crawl links rendered only in JavaScript. It's not as sophisticated as Googlebot!

  • Pages longer than 10000 pixels will get cropped to 10000px long. (You can manually modify Chromy to extend this to 16384px but Chrome headless itself refuses to go any longer than that.)

Tips

  • Each site's files has its own folder under sites, e.g. sites/examplesite_com_au.

  • To approve only selected test results, run --test -u url1 url2 ... -o examplesite.com.au and then --approve examplesite.com.au. (The -o option overrides the site directory to be examplesite_com_au even if your first test URL is different e.g. info.examplesite.com.au).

  • The HTML for every crawled URL is copied to the html directory under the directory for your site, e.g. sites/examplesite_com_au/html. This is not used by Snapsite, it is just there so you search over the whole site's content if you want to.

  • To delete a stubborn folder on Windows (filenames too long?), use the --delete action.

  • You can hit ^C in the middle of a --crawl, run a --reference, continue the crawl by running --crawl again, and then --reference again to only pick up the newly crawled pages.

  • When running --crawl again, its default behaviour is to resume the last crawl.

  • Use --crawl -f to suppress this resume function, and force a fresh crawl from scratch. This will erase the whole site directory and start over.

  • Similarly, running --reference again will crawl anything new in crawled-urls.log (it compares to referenced-urls.log).

  • Use --reference -f to force a fresh new reference — this erases all existing reference and test files but keeps crawled-urls.log.

  • --test always runs against every URL in referenced-urls.log.

  • Site stability: if your site's appearance naturally varies a lot each time you look at it, you will see a lot of failures. You can mitigate this somewhat by modifying your site-specific sites/<site>/config/onReady.js to cover-up or remove elements which keep messing up your results uselessly (see the coverVideo() example for the Toyota website).

  • Use --reference -u url1 url2 ... to reference an ad-hoc list of URLs on the command line. The site directory will be named after the domain name of the first URL. These URLs will be added to referenced-urls.log so that any subsequent --test will pick them up.

  • Use --test -u url1 url2 ... to test an ad-hoc list of URLs. These URLs will be added to tested-urls.log.

Built with

  • Supercrawler - Spiders the website
  • BackstopJS - Captures website screenshots from URLs
  • SQLite - Supercrawler needs to store its crawl progress somewhere, and this is that somewhere

Versioning

Use SemVer for versioning. For the versions available, see the tags on this repository.

Who do I talk to?

Please contact the following people for additional information: