https://discord.cloudflare.com logo
Join Discord
Powered by
# workers-discussions
  • j
    Do you have an example of what you're trying to do exactly? I don't have anything directly to throw you, but take a look at the 
    text
    handler in my post at https://jross.me/using-cloudflare-workers-htmlrewriter-to-extend-ghost-pro/#cloudflare-workers, and how I do the 
    lastInTextNode
    stuff to get the full text of the entire element before doing my replacements.
  • g
    Copy code
    html
<div>abc<p>middle</p>xyz</div>
    I can't figure out a way to see that 
    abc
    and 
    xyz
    are siblings.
  • g
    element
    gets called whenever we go up into a node, but 
    text.lastInTextNode
    isn't exclusively going down a node. In the example above, 
    abc
    is also 
    lastInTextNode
    .
  • j
    Hmm yeah. Due to the streaming nature of HTMLRewriter, I don't think there's a way to really know that unfortunately. I can't think of a way at least. https://up.jross.me/8lxkfas9
  • g
    Alright, thanks for looking 🙂
  • j
    https://community.cloudflare.com/t/non-streaming-htmlrewriter-response/119941 may be of interest. Similar-ish use-case with wanting to scan an entire element before doing anything with it.
  • g
    It's a bit above me, but I think that's only working because we know that a 
    body
    exists and will definitely follow the 
    head
    ?
  • g
    If I'm just parsing arbitrary HTML (inside a body), then there's no guarantee on any of the elements.
  • j
    Yeah. Doing it generically for an entire document would be crazy. This works because there's at least some constraints with what you're expecting and looking for.
  • j
    If no one has another solution, I guess the best one would be just finding a JS dom library and doing it synchronously after the document has been fetched
  • g
    Yup, I'm just setting up parse5 now instead.
  • g
    Honestly, no biggie. I just got obsessed with trying to make it work with HTMLRewriter. At least I learnt a bit more about how that works lol
  • j
    Yeah I dived down a similar rabbit hole in that blog post with text chunks and getting that working as expected. Really cool stuff, within some constraints
  • g

    https://i.imgur.com/g24X9RI.pngâ–¾

  • g
    Urgh
  • e
    Just to recap: you want to extract text from certain elements? But 
    lastInTextNode
    also shows up before 
    <p>
    so it makes it hard to get the whole text?
  • g
    Yes. So as far as I can tell, it's impossible to differentiate between:
    Copy code
    html
<div>abc</div><div>xyz</div>
    and
    Copy code
    html
<div>abc<div>xyz</div></div>
    with HTMLRewriter, because they all come out as
    Copy code
    element: div
text: abc (lastInTextNode)
element: div
text: xyz (lastInTextNode)
    once you combine any text chunks (as you're supposed to)
  • x
    Hello, would anyone have recommendations about storing time series from Workers please? There's some data I'd like to save about every execution (aka. click) that'd allow me to do some stats and graphs. I've been unable to find much information, I don't believe Workers KV is powerful enough and it can also become pricy depending on how many clicks there are of course. I ignore if Workers DO would be a good fit either, though at the same time I might be missing something like "it's already included somewhere" and you only need to point Grafana there? Thanks in advance to anyone who has any idea, and whoever reads this… have a nice day! :)
  • a
    oh hey xenthys
  • x
    oh hey advaith, it's been a while, I've been seeing you contribute a lot to the datamining repo
  • j
    I haven't seen many people attempting time series
  • x
    I would've thought time series are used everywhere, what are people doing otherwise :c
  • j
    Copy code
    +------------+
                       |            |
                       | Aggregator |
                       |            |
                       +-----+------+
                             ^
                             |
+--------------+       +-----+------+       +--------------+
|              |       |            |       |              |
|    Client    +------>+   Worker   +------>+    Origin    |
|              |       |            |       |              |
+--------------+       +------------+       +--------------+
  • j
    something like that
  • j
    I'd assume would work
  • j
    but you are still reliant on an external aggregator of some sort to store & query that
  • j
    you are constrained by CPU time, which means anything intensive isn't going to be that easy (yet, #773219397422284820 will make it easier)
  • x
    CPU time shouldn't be an issue to send some stats over the wire, it's not active computing
  • j
    yeah
  • x
    and yeah it'd work, but aggregate with what and where remains the question, if I'm gonna depend on an external database hosted on a few machines that's gonna make Workers lose a part of their advantage of being an edge computing service
1...585960...2509Latest