A names index for a search box

日本語

@johnmorrisdotca/hoshi-stars/names is the list a search box needs: every one of the 5,080 bright stars and the 110 Messier objects as a row, built from the tables the package already holds (no new data), so a page pays for it only when it searches. The 11,788 other NGC and IC objects are @johnmorrisdotca/hoshi-stars/names/ngc-ic, a chunk of their own.

import { findName, nameRowById, nameRows } from "@johnmorrisdotca/hoshi-stars/names";
import { findNgcIc } from "@johnmorrisdotca/hoshi-stars/names/ngc-ic";

const star = findName("alpha Ori");
console.log(star.id, star.name.en, star.name.ja, star.pick);

const galaxy = nameRowById("m31");
console.log(galaxy.id, galaxy.name.en, galaxy.kind, galaxy.pick);

const nebula = findNgcIc("North America Nebula");
console.log(nebula.id, nebula.name.en, nebula.kind);

const rows = nameRows();
console.log(`${rows.length} rows, first ${rows[0].id}`);

betelgeuse Betelgeuse ベテルギウス S:2061
m31 Andromeda Galaxy galaxy D:M31
ngc7000 North America Nebula nebula
5190 rows, first sirius

The output is:

betelgeuse Betelgeuse ベテルギウス S:2061
m31 Andromeda Galaxy galaxy D:M31
ngc7000 North America Nebula nebula
5190 rows, first sirius

A row

Finding one

findName(text) returns the one row a text names exactly, ignoring case, spacing, accents and how the Greek letter is spelt (α, alpha, Alp, alf, a superscript π³ or a number written apart pi 3 Ori): Betelgeuse, alpha Ori, α Orionis, HR 2061, M 31, Messier 31, NGC 224, Andromeda Galaxy, ベテルギウス. A text that two rows share (Alpha Centauri is both α1 Cen and α2 Cen) finds none. searchNames(text, { limit, kinds }) lists the rows whose names start with the text and then those that contain it, each by weight, for a type-ahead box. The NGC and IC chunk has the same three: ngcIcRows, ngcIcRowById, findNgcIc and searchNgcIc.

A small first chunk

A search box wants about 20 kB first, not 135. @johnmorrisdotca/hoshi-stars/names/named holds only the 335 stars that have a common name and the 110 Messier objects (445 rows), with nameRows, nameRowById, findName and searchNames that work on them at once. Each row is the same row, with the same id, names, designations and position, that /names gives for that id, so a page can show results from this chunk and load /names (and /names/ngc-ic) afterwards for the rest without an id changing.

import { findName, nameRows, searchNames } from "@johnmorrisdotca/hoshi-stars/names/named";

const rows = nameRows();
console.log(`${rows.length} rows, first ${rows[0].id}`);
console.log(findName("alpha Ori").id, findName("NGC 224").id);
console.log(searchNames("andro", { limit: 3 }).map((row) => row.id).join(" "));

445 rows, first sirius
betelgeuse m31
m31 alpheratz mirach

The output is:

445 rows, first sirius
betelgeuse m31
m31 alpheratz mirach

The 335 stars come from a small table that scripts/make-names-named.ts writes from the bright-star table (the Yale data already credited in NOTICE.md; no new data). The ids are settled over the whole table when it is written, so the chunk needs no star table, and a test runs the script again and fails if the table on disk differs. /names/named measures 22.7 kB gzipped. A star without a common name (alpha-centauri, hr-1234) is not in it: a search that finds nothing there goes on to /names.

Your own index

The pieces the three entry points are made of are exported too, for a page that keeps its own list of rows (the rows of one catalogue, or a short list for one page). They are in /names and /names/named, and all but deepSkyRow in /names/ngc-ic.

• NAME_KINDS is the list of kinds a row can have (star, cluster, nebula, galaxy, other), in the order the type NameKind gives them.
• nameKey(text) is the text folded the way findName compares: lower case, no accents, the Greek letter spelt out, one space, a superscript as its digit. Messier 31, M 31 and m31 all give m31, and NGC 0224 gives ngc224.
• slugOf(text) is an address-safe id from a text: lower-case letters and digits in runs joined by one hyphen, an accent dropped (Boötes gives bootes) and a Greek letter spelt out. It is what the ids of the rows are made with.
• deepSkyRow(object, constellations) is the row of a deep-sky object from /deep-sky or /deep-sky/ngc-ic, with the constellations by abbreviation for the names of the one it lies in.
• nameIndex(rows) is the index over a list of rows: rows(kinds), byId(id), find(text) and search(text, { limit, kinds }) work as nameRows, nameRowById, findName and searchNames do. The rows come out largest weight first, and it throws if two rows have the same id.

import { constellations } from "@johnmorrisdotca/hoshi-stars/constellations";
import { deepSkyObject } from "@johnmorrisdotca/hoshi-stars/deep-sky";
import { NAME_KINDS, deepSkyRow, nameIndex, nameKey, slugOf } from "@johnmorrisdotca/hoshi-stars/names";

console.log(NAME_KINDS.join(" ")); // star cluster nebula galaxy other
console.log(nameKey("Messier 31"), nameKey("α Ori"), nameKey("NGC 0224")); // m31 alpha ori ngc224
console.log(slugOf("Rigil Kentaurus"), slugOf("Boötes")); // rigil-kentaurus bootes

const byAbbreviation = new Map(constellations().map((c) => [c.abbreviation, c]));
const row = deepSkyRow(deepSkyObject("M31"), byAbbreviation);
console.log(row.id, row.kind, row.pick); // m31 galaxy D:M31

const mine = nameIndex([row]);
console.log(mine.find("Messier 31").id, mine.rows(["star"]).length); // m31 0

star cluster nebula galaxy other
m31 alpha ori ngc224
rigil-kentaurus bootes
m31 galaxy D:M31
m31 0

Size

The rows are made when first asked for from the tables the package already ships, so the code adds a few kilobytes: /names measures 135.7 kB gzipped (the star table with the Messier table) and /names/ngc-ic 226.4 kB (the large deep-sky table, and nothing from the stars). Both are over 60 kB, so they are chunks, with the 22.7 kB /names/named in front of them, and a page that only wants stars and Messier objects never loads the NGC and IC table. Written out as JSON the rows are larger than the tables they come from, because every spelling of a name is spelt: of the 5,190 rows of /names about 300 kB gzipped (5,080 stars about 296 kB), and the 11,788 of /names/ngc-ic about 332 kB (the galaxies are most of it). A site that sends its own copy should send the fields it searches and shows, not whole rows.

Known gaps

• No HD numbers for the bright stars. HD 39801 does not find Betelgeuse: the package holds a Henry Draper number only for the stars of /nearby. The Bright Star Catalogue has the column; adding it is a data re-run that needs John's yes (ticket MODULES-1224).
• Flamsteed numbers of a star that also has a Bayer letter (58 Ori for Betelgeuse) are not in the data the package exposes, so they are not typed names; a star with only a Flamsteed number has one.
• The 5,565 nearby stars are not rows yet.