Is there a way to removeabortReason and _urlRewrite from the serialized page Context?

[ChristophP]

Hi Rom,

our SEO has recently complained about our HTML that is quite large.
There are multiple contributing factors but one angle we are aiming at is to remove things from Vike page Context that gets serialized into the HTML.

We noticed that there are things in the pageContext such as abortReason, _urlRewrite after server side redirects.
Is there a way to remove these? We only use server routing.

Some more background: _urlRewrite contains urls, that is the SEO is worried will get indexed (which I'm not sure is actually happening but apparently there are cases where google crawls JSON content).

PS: I brought up the issue of sponsoring internally. I didn't realize it had ended, which is likely due to the companies policy to renew it every 6 months.

[brillout]

I ain't sure the few superfluous pageContext properties passed to the client make much of a difference — I guess you do SSR so I'd expect the page content rendered to HTML to be a lot larger. For example, AFAICT, pageContext.abortReason is a drop in the ocean. (In general, I'm always careful to minimize as much as possible what is sent to the client-side and when I do send any superfluous bit it's typically a conscience trade-off.)

But definitely let me know if you find something superfluous that is relatively large.

apparently there are cases where google crawls JSON content

Interesting. Can you elaborate on that? That's quite surprising — and something Vike should be aware of it it's indeed the case.

PS: I brought up the issue of sponsoring internally. I didn't realize it had ended, which is likely due to the companies policy to renew it every 6 months.

Thank you 💚

[ChristophP]

One way to do it is to be able to add custom types to https://github.com/brillout/json-serializer (this is what Vike uses to serialize JSON). I'll have to look if it isn't too complex — @brillout/json-serializer aims to be lightweight.

That sounds like a good appoach. Thanks for looking into it.

[brillout]

With 0.4.235-commit-8f42b16 (#2542):

- <script id="vike_pageContext" type="application/json">{"pageId":"/pages/index"}</script>
+ <script id="vike_pageContext" type="application/json">{"pageId":"!/pages/index"}</script>

This change applies to all JSON values (including _urlRewrite and nested ones).

I presume this will prevent Google from treating filesystem paths found in JSON as a URL.

Let me know if it works.

[ChristophP]

Great thank you for getting at this.
One question: Did you find any resources about that ! is enough to prevent Google from indexing it?
My next task to update Vike (looking forward to be on a current version).
Will report back about the Google indexing, which could be a few days until I hear back from the SEO team.

[brillout]

No I'm just guessing and I ain't 100% sure, so we'll have to test this. Let's see.

(FYI I tried other more "bullet proof" approaches but they turned out to be quite complex to implement or having other kinds of drawbacks.)

👍 Keep me updated.

[ChristophP]

Ok we'll see. I'll check back with the SEO team. One thing they suggested before war to turn /some/url/path into \/some\/url\/path. But maybe the ! ist enough as is.

[ChristophP]

Hi Rom, reporting back.
So apparently there's two additional things.

1. Prepending ! is not enough, because Google seems to also consider urls as substrings.
2. Because if this the _abortCall property is also relevant, which doesn't get masket currently ("render(\"/some/url\")")

So I propose the following fix for all props that might contain a url somewhere in the string:
Doing a .replaceAll('/', '\\/') on them.

Current workaround. Is for us to manually modify the pageContext this way.

[brillout]

👍 Thank you for reporting back. I'll have a look at it.

[brillout]

Done (#2603) and pre-released as 0.4.236-commit-eab75f6.

Let me know how it goes.

[ChristophP]

Great, will update soon and let you know.