> Because the main goal is to use it for more complex variables, like 
> the about-to-be-added BillingAddress type. This example was just for 
> illustration.

Ack.

> I don't see why this would be hacky? The tutorials on graphql.org for 
> example consistently use variables even for simple queries. It's one 
> of the fundamental mechanisms of GraphQL.

Hacky isn't the right word. Clunky maybe is better. As you can see I'm a 
sucker for technical terminology.

I think Drew touched on it better in another email. The interface feels

Just for discussion reasons, any reason you don't create the query like 
so:

    query user($username: String = "bitfehler") {
	userByName(username: $username) {
	  canonicalName
	}
    }

We attempted to do something similar with variables for our GraphQL 
playground (which is a total copy of srht's) but it felt hacky using it.

We scrapped it in the end as the QL supports variable declaration like 
the example above.

I ran an into an issue with the new implementation where a module is
used to start queue(s) but knows nothing about the current application
(as far as queue size, workers, etc. go).

Being able to get the given size allows us to use a formula to
dynamically generate the number of workers to use when starting each
queue.
---
 queue.go | 25 ++++++++++++++++---------
 1 file changed, 16 insertions(+), 9 deletions(-)

diff --git a/queue.go b/queue.go
index 681129e..7f16dd1 100644
--- a/queue.go
[message trimmed]

The slices package was introduced in Go 1.21. If we're going to use this 
package then go.mod needs to be updated as it's currently coded to 1.18.

Peter

> Frequently I want to use the `:open-link` command, but the email has a 
> dozen
> links that all look vaguely similar, and I want to just search for the unique
> bit (usually 'unsubscribe'), but the command palette matching seems to be
> prefix-based, which isn't useful in this context.

For this I simply map C-l to use urlscan[0] instead of the default 
:open-link functionality. It is easy to decide which link to click based 
on where the link is assigned in the email. For me this has always been 
easier to track for emails with a ton of links that are all basically 
the same, just a few characters off to determine which link was clicked.

> - Should this be only for `:open-link`, for all command arg 
> completion, or configurable?

>Thanks you Peter, it worked!

Glad that resolved it.

>Have a great day.

You as well.

Peter

On 04/09, thomas wrote:
>I know, the issue I have is that when I press "Authorize account 
>access", a request is made to `meta.local/oauth/authorize`, which so 
>far is fine, and then I got a 302 from `meta.local` to `http://git.local/?exchange=xxxxxxxxxxxx&scopes=profile%3Aread%2Ckeys%3Aread&state=%2F%3F`. 
>And basically at that point I'm back to `git.local` but 
>unauthenticated. I feel like I'm living in the Groundhog Day movie! :)

I believe there is a process to setup oauth clients and trust them by
manually updating the db entries for each service. It should be in the
docs but it's been years since I setup my instance so I don't remember
exactly.

I believe you make the oauth clients for each service (git, man) and
then something like `update oauth_clients set trusted=true where id=XXX`

On 04/09, thomas wrote:
>I'm trying to self-host sr.ht on my local machine. So far, I've 
>managed to set up man and git.
>
>However, when I'm trying to log-in into git, I see the oauth page 
>("Authorize account access") but "Proceed and grant access" doesn't 
>work. Basically it redirects me to git.local/?exchange=xxxxxxxxxx&scopes=profile%3Aread%2Ckeys%3Aread&state=%2F%3F

Hi Thomas. The `meta` service is a requirement for all other services.
So no matter how few of the services you want to run, they all depend on
meta. That's where you'll do the oauth setup, etc.

Peter

# HG changeset patch
# User Peter Sanchez <peter@netlandish.com>
# Date 1712515681 21600
#      Sun Apr 07 12:48:01 2024 -0600
# Node ID 4a8e7f8c497f1c1caba34c3343e60fd81850d49a
# Parent  d9fc0f61ffddf26171417868763212d46a9be0c9
Updating hgsrht-keys to use latest scm.sr.ht/srht-keys to be consistent

diff --git a/hgsrht-keys/go.mod b/hgsrht-keys/go.mod
--- a/hgsrht-keys/go.mod
+++ b/hgsrht-keys/go.mod
@@ -3,8 +3,8 @@ module hg.sr.ht/~sircmpwn/hg.sr.ht/hgsrh
 go 1.13
 
[message trimmed]

Just a follow up that this indeed appears to have resolved the issue.
I should have thought of the celery brokers.

Thanks again!

Peter

-- 
Peter Sanchez
pjs@petersanchez.com   https://petersanchez.com
P.S.  What I'm up to now: https://petersanchez.com/now/