Partial Labels

[hynek]

Unless I’m missing something, it’s currently necessary to specify all labels at once. It would be nice tho if I could do something like this:

M = Counter("a counter", "counter that counts", ["id", "type"])
...
c = M.labels({"id": id})  # id is determined at runtime/initialization
...
c.labels({"type": "internal").inc()  # type is unclear until it happens
...
c.labels({"type": "external").inc()

Does that make sense?

[brian-brazil]

I think such an API would cause confusion as it's not clear whether a given labels use is complete.

I'd expect that you'd always know all the labels in play when incrementing, so this isn't going to buy you much. I'm also considering dropping the dict argument, in favour of kwargs.

[hynek]

Well, I do know, but if one of three labels is dynamic, it’s still pretty painful to always type them out every single time. Copy and paste is error-prone and annoying.

And yes, kwargs would be great, because currently the “best” way I found to precompute labels is:

    def request_time_for(self, type):
        return self._request_time.labels(*(self._labels + (type,)))

(can’t use mutable dicts for I’d have to copy them first…)

---

Anyhow, if you wanna keep the official client as minimal as possible I can stop opening bugs and just build my own convenience layer called “lighter” or something. :D

I just found the current behavior highly discouraging for using labels, that’s all.

[brian-brazil]

Just thinking there the kwargs would help you as you could do

c.labels(type="internal", **standard_labels).inc()

[hynek]

yeah, that’s what I meant by “would be great” :)

[raul1991]

I want to build dynamic labels with this client lib. I have the following code

self.online_time_span = Gauge('online_time_span', 'How long someone has been online', ['value'])

self.online_time_span.labels('online_time').set(time_span)

In the above statement I need the labels to represent something like

'online_time_spent_<USERNAME_WHICH_UNIQUE_COMES_FROM_API>'

Examples:

'online_time_spent_John'
'online_time_spent_Maria'
'online_time_spent_Foo'

Is it possible ?