Skip to content

cleanup: highlight country once we have contributors there - #57

Merged
InftyAI-Agent merged 6 commits into
InftyAI:mainfrom
kerthcet:chore/map-coauthors
Sep 11, 2026
Merged

cleanup: highlight country once we have contributors there#57
InftyAI-Agent merged 6 commits into
InftyAI:mainfrom
kerthcet:chore/map-coauthors

Conversation

@kerthcet

Copy link
Copy Markdown
Member

No description provided.

kerthcet and others added 6 commits September 11, 2026 13:43
ky-hiverge belongs on the map — Cambridge, like kerthcet — and was not on it. The
account has no commit and no pull request of its own in these repositories: what
it has is five commits written with somebody else, where GitHub attributes the
commit to one author and names the second in a `Co-authored-by:` trailer in the
message. The trailer is in no field of the API's response, and this script read
`author` and nothing else, so the second person was read as nobody.

Not a corner case. Every project added to Awesome-LLMOps arrives as a pull
request that inftyai-agent squash-merges, which makes that machine account the
author — excluded here as a bot — and the person who did the work the co-author.
So that work counted for nobody, while the same work in a repository whose merges
a human presses counted in full, and about twenty of the people already on the
map are Awesome-LLMOps submitters. This is one rule applied to both, not a new
definition of contributor.

A login is read from the trailer's address and from nowhere else, and only from
`id+login@users.noreply.github.com`, the one form that names an account. The
display name beside it is free text: `Kerry He` is `kerry-he` and looks
guessable, `Se7en <chengzw258@163.com>` is `cr7258` and is not, and a name that
happens to be a stranger's login would credit the stranger. The rest are
reported at the end of the run — every one of them turned out to be somebody an
ordinary commit or a noreply trailer elsewhere had already found. Logins are
counted case-insensitively for the same reason: `googs1025` from the API and
`86391540+googs1025@users.noreply` from a trailer were two contributors before.
Three logins now 404 (renamed or deleted since their commit); they are skipped
with a warning rather than counted, since there is no profile to place them by.

53 contributors becomes 72, 36 placed becomes 47, and the map goes from 23 marks
to 29: London and Kuala Lumpur and Cleveland and Tampa and Romania and Italy
appear, Bengaluru and Cambridge and Shanghai grow. Cambridge now reads
`kerthcet, ky-hiverge`.

One loss worth saying out loud: pranshuchittora writes "Bengaluru, IN", so with
that alias learned there is a city in India, and a city prunes its country's
fallback mark. SanjanShiv and VedantMadane, who say only "India", are now in the
Asia share and on no card — the same trade the United States mark already makes.

The six new dots displaced eleven labels, which are hand-placed again here: no
label crosses another, sits on a dot that is not its own, or leaves the drawing,
measured from the rendered page at 1440 and checked by eye at every mark. Paris
now sits in the 15-unit gap between Berlin's dot and Italy's, London and
Amsterdam stack above a cluster of five dots that no longer has room beside it,
and New Zealand hangs below-left of its dot so that it stops reading as a second
label for Melbourne's.
A country-level mark was a fallback that lost its place to the first city in the
same country: "a blob over China beside five Chinese cities reads as a sixth city
that nobody lives in", which is a fair thing to say about the blob and the wrong
thing to weigh against. The mark was never the only thing dropped. The people it
stood for went with it — off the card that names them, into a region percentage —
so writing "India" rather than "Bengaluru" was enough to be counted and not
shown. Four of the forty-seven placed contributors were in that position: awdemos
and twinsgeeks under the United States, SanjanShiv and VedantMadane under India.

The contributors are the baseline, so nothing that has somebody under it is
dropped for the drawing's sake. 29 marks becomes 31, and the counts on the marks
now add up to `located` exactly — 47 of 47 people on a card, where it was 43.
Nothing else moves: the shares were always computed from every matched place
rather than from the surviving ones, so the legend is unchanged to the point.

What is left of the blob problem is a label, and the label is the answer to it.
"India" beside a dot in the middle of India is not a city, and the card it opens
lists the two people who wrote it that way. Four labels are re-placed around the
two new dots: India to the left of its own dot with Bengaluru moved below its
own, United States over Nevada with its dot 11 units to the right, and San
Francisco out over the Pacific, since its old spot above the dot ran into the new
one in Kansas.

The report of orphaned contributors goes with the pruning — it existed to make
this loss visible, and there is no longer such a thing.

Measured after: 31 targets over 31 dots, each centred on its own to within
0.06px, every card opening on top with both avatars loaded and none crossing a
viewport edge at 320, 390, 768 or 1440, and all closed again at rest. No label
crosses another, sits on a dot that is not its own, or leaves the drawing.
The generator's report of locations it cannot place named one: "limassol",
which is a city in Cyprus and now an entry with the rest. Cyprus is filed
under Asia, which is where the UN geoscheme puts it and where Turkey and
Israel already are in the region table.

Refreshed with it: 48 of 75 contributors placed, in 27 cities across 20
countries, on 33 marks. Two are new since the last run — South Africa and
Limassol — and the report is clean apart from three names with no GitHub
account behind them, which nothing here can look a location up for.

Cambridge drops from two contributors to one: ky-hiverge has cleared the
Location field on their profile, and an empty location is not something
this can place. They are still counted among the 75.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Shanghai gained a contributor in the refresh, which is 28px more card: 249px
against a 109px map at 320px wide, hanging 56px past its bottom edge where
the 3.5rem gap covered 42px. The card's edge was back over the legend's
"Asia" row by 6px.

4.5rem, measured the same way — 9px above the legend's box — and the comment
now carries what moves it: 12px per face, 7rem for a full card of ten. That
is spacing a 109px map cannot carry, so the note says what has to give if it
comes to it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The map had one kind of mark for two different claims. A contributor who gave
"San Francisco" and one who gave "United States" both became a dot, and the
second one landed on a national centroid — 39.5N 98.35W, a field in Kansas —
drawn exactly like a city 500 units away that is genuinely known.

So the two claims are now drawn differently: every country anybody is in is
filled whole, and the cities inside it keep their dots on top. The shade says
somewhere in here, the dot says here.

tools/worldmap.mjs takes its geometry from world-atlas' countries-110m.json
instead of land-110m.json. Same coastline, and this file carries the land
silhouette and the per-country outlines built from one shared set of arcs —
two files would have simplified the same coast twice, and disagreements of a
pixel would have shown as a fringe of unshaded land along every shaded coast.
It emits an outline for every country tools/locations.json can place anybody
in, not the countries on the map today: data/contributors.json is regenerated
on every contributor change, and this file would otherwise need rebuilding
over the network to catch up. Fifty-three outlines, 19 KB in a file nothing
downloads; the page carries the twenty that are used, which is 10 KB of path.

Three things follow for a country-level mark. Its dot goes, and so does its
pulse — a ring growing out of open country with no mark at its centre reads as
an artefact. And its name is only written where the country is the only thing
there: "United States" beside San Francisco, Palo Alto, Los Angeles, Chicago,
Cleveland and Tampa was a label for the shape those six dots already stand on,
and it was spending room in the most crowded corner of the map. Romania, Italy,
South Africa and New Zealand — the four countries with no city on the map —
keep theirs. The marks themselves stay: those contributors gave a country and
no city, so the card that opens on the country's centre is the only place their
names appear.

The shade is a darker neutral, not a tint of the mark's colour, which is where
this started at 35% alpha. The trouble was never the strength: North America,
China, India and Australia are all shaded, so the largest warm shapes on the
map were the ones carrying the vaguest information. $map-region is the land's
own hue at 84% lightness against its 93%, so the dots stay the only colour on
the map. 86% was tried first and lost Romania, Italy, Ireland and Cyprus
entirely, which is where the shade is the only thing saying anybody is there.

Each country also takes the hairline of page background the dots use, for the
same reason the dots have it: without it the United States and Canada fused
into one mass from the Arctic to Mexico, China and India into another, and
eight European countries into a blob — a map saying "this continent" where the
data says which countries. 0.8 units against the dots' 2.5, because this is a
border and not a keyline.
Signed-off-by: kerthcet <kerthcet@gmail.com>
@vercel

vercel Bot commented Sep 11, 2026

Copy link
Copy Markdown

@kerthcet is attempting to deploy a commit to the InftyAI Team on Vercel.

A member of the Team first needs to authorize it.

@kerthcet

Copy link
Copy Markdown
Member Author

/lgtm
/kind cleanup

@InftyAI-Agent InftyAI-Agent added needs-triage Indicates an issue or PR lacks a label and requires one. needs-priority Indicates a PR lacks a label and requires one. do-not-merge/needs-kind Indicates a PR lacks a label and requires one. approved Indicates a PR has been approved by an approver from all required OWNERS files. labels Sep 11, 2026
@vercel

vercel Bot commented Sep 11, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
website Ready Ready Preview Sep 11, 2026 3:09pm UTC

@InftyAI-Agent InftyAI-Agent added lgtm Looks good to me, indicates that a PR is ready to be merged. cleanup Categorizes issue or PR as related to cleaning up code, process, or technical debt. and removed do-not-merge/needs-kind Indicates a PR lacks a label and requires one. labels Sep 11, 2026
@InftyAI-Agent
InftyAI-Agent merged commit a8f6a61 into InftyAI:main Sep 11, 2026
40 of 42 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

approved Indicates a PR has been approved by an approver from all required OWNERS files. cleanup Categorizes issue or PR as related to cleaning up code, process, or technical debt. lgtm Looks good to me, indicates that a PR is ready to be merged. needs-priority Indicates a PR lacks a label and requires one. needs-triage Indicates an issue or PR lacks a label and requires one.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants