Repository navigation
Installation instruction ambiguities #304
Description
The README file seems to be ambiguous with respect to:
- Whether building the auditor is a prerequisite to all modes of use.
- What Grunt installation steps are required before
grunt build. Apparently local installation of Grunt and creation of a Gruntfile are required, but this is not stated, nor is the content of the necessery Gruntfile specified. - Whether the
puppeteerscript could be adapted to useplaywright. - How the output from
HTMLCS_RUNNER.run()can be captured for further processing in either thepuppeteeror thejsdomcase.
Activity
guygastineau commentedon Feb 17, 2023
That is currently my issue. I am using the script with both marionette and webdriver (using geckodriver). It runs, but I get
nullas the return value. Is it printing it to the console or something?I discovered this script is what pa11y uses to scan for accessibility. Maybe I can figure out how they are getting the information out of it.
guygastineau commentedon Feb 17, 2023
I found the code in
Contrib/PhantomJS/runner.js. It is indeed printing the results to the console separated simply with vertical bars|. TheHTMLCS_Run.jsfile in the same directory has a function that suggests it returns JSON, but I don't know how to use that. TheHTMLCS_RUNNER.runfunction says it takes a callback, but it doesn't use it. I was hoping I could just pass some callback to turn it all into JSON or something. Anyway, it looks like I'll have to patch it, since I don't find tricks for capturing the console output suitable for my needs.jrpool commentedon Feb 18, 2023
Testaro (https://www.npmjs.com/package/testaro) runs HTMLCS (and other packages). Testaro captures the output of HTMLCS and integrates it into its own output. The code and README file of Testaro show how it does that.
guygastineau commentedon Feb 19, 2023
@jrpool thank you for the link. Hopefully, I didn't take this issue too far off topic.
I wrote a new runner locally that actually returns a value, but then I realized the machinery at the heart of HTMLCS is like continuation passing style but without return values. I started adding
returnsto the final expressions in all branches, but then work was over and I started drinking to cope with having to walk the graph manually.I noticed many functions return false under some failure conditions, but otherwise they implicitly return null. I just want to get return values from the callbacks threaded through HTMLCS. Do you think a PR adding this while preserving backwards compatibility would be welcome? I'm sure I'm off topic now.