In order to do all this, the developer needs to understand what the DOM looks like on the Hypothesis page, and what the format of the Arxiv API results XML looks like, in order to get the parsing correct. It involves parsing the DOM on the Hypothesis page to find references to Arxiv, querying the Arxiv API (but only after dividing the IDs into batches), parsing the XML results returned by the API, and again parsing and updating the DOM on the Hypothesis page to update the results.
This is not a trivial script by any stretch of the imagination. I use Chrome, so I would need a Tampermonkey script. Tampermonkey is the equivalent for Chrome. Writing a Greasemonkey script is time consuming and it wouldn’t be worthwhile for such a small usability improvement. Their exact suggestion was: You can write a Greasemonkey script which parses the Arxiv PDF links on Hypothesis pages, gathers their ArxivID, queries the Arxiv ATOM API with one HTTP call (using an aggregated list of ArxivIDs as a search parameter) and updates Arxiv PDF links titles on Hypothesis pages.Įxample of an API query with multiple ArxivIDs: Īt first I thought this wasn’t such a helpful suggestion. Someone suggested I could write a Greasemonkey script to do this. This would allow me to replace the useless IDs with the more useful titles on pdfs whenever I take notes on them. So, I made a feature request on GitHub, asking if the developers of hypothesis could make the titles of pages editable.
This makes it hard to find what I’m looking for. It’s a great tool, but it has a limitation: when browsing my notes, it only shows the ID of the pdf, not the name of the PDF. I use it to take notes on machine learning pdfs on Arxiv. For context, Hypothesis is a tool for taking notes on the web, including on PDFs. I am blown away by GPT-4’s programming capabilities.
It took well under half an hour, which is likely many times faster than it would have taken me alone.
0 kommentar(er)
