!SDKYnTgYyOXNBpJySp:matrix.org

cylc protocols & APIs

8 Members
HTTPS, WebSocket, ZeroMQ; REST, GraphQL1 Servers

Load older messages


Timestamp Message
18 Dec 2018
07:31:57@davidmatthews:matrix.orgDavid Matthews joined the room.
09:39:35@wxtim:matrix.orgwxtim joined the room.
09:48:36@matthewrmshin:matrix.orgMatt Shin I'd vote for having a single meta dict. 1. It reflects the data structure in suite.rc. 2. It looks less complicated. 3. https://github.com/cylc/cylc/pull/2387 is essentially what we have done internally to make this a single dict.
09:55:58@dwsutherland:matrix.orgDavid SutherlandYes, that would also be the easiest to implement, conceptually the most tidy, and consistent with the suite/cylc. However, the catch is; you cannot select specific fields (i.e. title & url only) in the query (you get the whole lot!).. But seeing as this won't need to be updated that often, it's probably not a big deal.
09:57:04@dwsutherland:matrix.orgDavid Sutherlandselect specific meta fields that is..
10:05:05@dwsutherland:matrix.orgDavid SutherlandThe single meta dict avoids the splitting of the meta dictionary into two also. Which sounds trivial, but if you are doing that for 1000 tasks X# of families then it would probably add up I suppose. For the global it's suite_meta = schd.config.cfg['meta'] globalql_data['meta'] = {x: suite_meta[x] for x in self.META_KEYS} g_meta_aux = {} for key in suite_meta: if key not in self.META_KEYS: g_meta_aux[key] = suite_meta[key] globalql_data['aux_meta'] = g_meta_aux
10:06:07@dwsutherland:matrix.orgDavid SutherlandWhich would reduce to globalql_data['meta'] = schd.config.cfg['meta']
10:08:00@dwsutherland:matrix.orgDavid Sutherland(or copy/deep_copy, but seeing as it's only a data read, then it should be fine)
10:09:05@matthewrmshin:matrix.orgMatt Shin It is probably OK, as it will be a one-off call. [meta] does not change during suite run time, (unless you reload the suite.rc).
10:10:10@dwsutherland:matrix.orgDavid SutherlandYes, exactly.
10:12:26@matthewrmshin:matrix.orgMatt Shin👍
10:28:01@dwsutherland:matrix.orgDavid Sutherland
"globalInfo": {
  "suite": "baz",
  "owner": "sutherlanddw",
  "meta": {
    "URL": "https://niwa-35595lvm.niwa.local:<port>/graphql",
    "group": "round table",
    "description": "Some suite baz",
    "title": "BAZ",
    "importance": "critical",
    "ranking": "one"
  },
  "reloading": false,
  "lastUpdated": 1545128661.219214,
  "status": "held",
  "runMode": "live",
  "newestRunaheadCyclePoint": null,
  "newestCyclePoint": "20170401T0000+13",
  "oldestCyclePoint": "20170101T0000+13",
  "treeDepth": 4
}
11:07:43@matthewrmshin:matrix.orgMatt ShinI wonder if it is better to separate the suite start up info (suite, owner, meta, run mode - info that can only be changed by start up or reload) from the suite run time info (the rest - info that can be updated by the main loop)?
11:07:59@matthewrmshin:matrix.orgMatt Shin(Sorry to be a pain.)
11:13:43@dwsutherland:matrix.orgDavid SutherlandNo I think that's a good idea; although you can do that by just specifying different fields in two queries instead of one. But the actual driving data should be split up in the backend in the same way, so the suite is only updating what it needs to within the loop
11:18:29@dwsutherland:matrix.orgDavid SutherlandIt's also an additional motivation (aside from the visualisation requirement) for the seperation of job and task (because the task holds dynamic information of the latest job)
11:20:01@dwsutherland:matrix.orgDavid SutherlandAnyway; off to sleep (^.^);
11:20:24@matthewrmshin:matrix.orgMatt Shin👍
20:39:05@kinow:matrix.orgBruno P. Kinoshita joined the room.
23:47:40@kinow:matrix.orgBruno P. KinoshitaFrom the UI point of view, there are no blockers in the suggestions here... unless there's the requirement for filtering using some meta field (which would also fall mainly on the graphql implementation, from the UI perspective it would be just a small change in some controls to send the right parameters to the backend). So 👍 for the latest comments so far 🙂
19 Dec 2018
00:04:11@kinow:matrix.orgBruno P. Kinoshita What would be more useful for the UI, is marking the fields that are definitely going to be present with that exclmation mark in the schema. So that we don't have to add a lot of if some var != null, etc. That'd be really great. e.g. task label, time point, suite name, etc
2 Jan 2019
16:13:12@oliver-sanders:matrix.orgOliver Sanders joined the room.
8 Jan 2019
02:33:02@revilo666:matrix.orgHilary Oliver Just wondering if new joiners can see earlier (pre-join) posts? Oliver Sanders you've just joined this room, can you see the page or two of existing posts?
07:21:41@davidmatthews:matrix.orgDavid Matthews Hilary Oliver: You can view the room anonymously (e.g in a private browsing window) back to when you made future room history visible to anyone
09:43:59@oliver-sanders:matrix.orgOliver SandersYep I can see right back to 2018-12-18#
16 Jan 2019
04:59:07@martinryan:matrix.orgmartin ryan changed their profile picture.
18 Jan 2019
10:14:48@oliver-sanders:matrix.orgOliver SandersI've started reading through the 0MQ docs, I might take a look at the publisher-subscriber relationship between the Suite and UI Server.
10:20:13@revilo666:matrix.orgHilary Oliver 👍
10:25:42@dwsutherland:matrix.orgDavid Sutherland Good - it'll be nice to unmask some of details that underpin our architecture
19 Jan 2019
08:06:42@kinow:matrix.orgBruno P. Kinoshita👍

There are no newer messages yet.


Back to Room List