131 lines
3.7 KiB
JavaScript
131 lines
3.7 KiB
JavaScript
import { test } from "node:test";
|
|
import { equal as eq } from "node:assert/strict";
|
|
|
|
import { CONTEXT_CHARS, reanchor, selectionSelectors } from "../../src/llm/web/focus_anchor.mjs";
|
|
|
|
test("a unique quote round-trips through its offsets", () => {
|
|
const text = "alpha we support the proposal beta";
|
|
const start = text.indexOf("we support the proposal");
|
|
const sel = selectionSelectors(text, start, start + "we support the proposal".length);
|
|
eq(sel.quote, "we support the proposal");
|
|
eq(sel.prefix, "alpha ");
|
|
eq(sel.suffix, " beta");
|
|
eq(sel.start_char, start);
|
|
const hit = reanchor(text, sel);
|
|
eq(hit.orphaned, false);
|
|
eq(hit.how, "position");
|
|
eq(hit.start, start);
|
|
eq(text.slice(hit.start, hit.end), sel.quote);
|
|
});
|
|
|
|
test("prefix and suffix stay within 32 characters", () => {
|
|
const text = "0123456789".repeat(6) + "QUOTE" + "abcdefghij".repeat(6);
|
|
const start = text.indexOf("QUOTE");
|
|
const sel = selectionSelectors(text, start, start + "QUOTE".length);
|
|
eq(CONTEXT_CHARS, 32);
|
|
eq(sel.prefix.length, 32);
|
|
eq(sel.suffix.length, 32);
|
|
eq(sel.prefix, text.slice(start - 32, start));
|
|
eq(sel.suffix, text.slice(start + 5, start + 5 + 32));
|
|
});
|
|
|
|
test("a repeated quote is disambiguated by prefix and suffix", () => {
|
|
const text = "before LEFT QUOTE after middle RIGHT QUOTE end";
|
|
const start = text.lastIndexOf("QUOTE");
|
|
const sel = selectionSelectors(text, start, start + "QUOTE".length);
|
|
const hit = reanchor(text, {
|
|
quote: "QUOTE",
|
|
prefix: sel.prefix,
|
|
suffix: sel.suffix,
|
|
start_char: null,
|
|
end_char: null,
|
|
});
|
|
eq(hit.orphaned, false);
|
|
eq(hit.how, "quote");
|
|
eq(hit.start, start);
|
|
eq(text.slice(hit.start, hit.end), "QUOTE");
|
|
});
|
|
|
|
test("stale offsets fall through to the quote", () => {
|
|
const text = "xxxx hello yyyy";
|
|
const hit = reanchor(text, {
|
|
quote: "hello",
|
|
prefix: "xxxx ",
|
|
suffix: " yyyy",
|
|
start_char: 0,
|
|
end_char: 5,
|
|
});
|
|
eq(hit.how, "quote");
|
|
eq(hit.start, text.indexOf("hello"));
|
|
});
|
|
|
|
test("whitespace differences still count as the quote", () => {
|
|
const text = "we\nsupport the proposal today";
|
|
const hit = reanchor(text, {
|
|
quote: "we support the proposal",
|
|
prefix: "",
|
|
suffix: " today",
|
|
start_char: null,
|
|
end_char: null,
|
|
});
|
|
eq(hit.orphaned, false);
|
|
eq(hit.how, "quote");
|
|
eq(text.slice(hit.start, hit.end).replace(/\s+/g, " "), "we support the proposal");
|
|
});
|
|
|
|
test("a slight text change is a fuzzy match", () => {
|
|
const text = "we suport the proposal today";
|
|
const hit = reanchor(text, {
|
|
quote: "we support the proposal",
|
|
prefix: "",
|
|
suffix: " today",
|
|
start_char: null,
|
|
end_char: null,
|
|
});
|
|
eq(hit.orphaned, false);
|
|
eq(hit.how, "fuzzy");
|
|
eq(text.slice(hit.start, hit.end), "we suport the proposal");
|
|
});
|
|
|
|
test("an absent quote is orphaned", () => {
|
|
const hit = reanchor("nothing here", {
|
|
quote: "missing sentence",
|
|
prefix: "no ",
|
|
suffix: " where",
|
|
start_char: 0,
|
|
end_char: 4,
|
|
});
|
|
eq(hit.orphaned, true);
|
|
eq(hit.how, "orphaned");
|
|
eq(hit.start, null);
|
|
eq(hit.end, null);
|
|
});
|
|
|
|
test("unicode quotes round-trip", () => {
|
|
const text = "The naïve café — 提案 stands.";
|
|
const quote = "naïve café — 提案";
|
|
const start = text.indexOf(quote);
|
|
const sel = selectionSelectors(text, start, start + quote.length);
|
|
eq(sel.quote, quote);
|
|
const hit = reanchor(text, {
|
|
quote,
|
|
prefix: sel.prefix,
|
|
suffix: sel.suffix,
|
|
start_char: null,
|
|
end_char: null,
|
|
});
|
|
eq(hit.orphaned, false);
|
|
eq(text.slice(hit.start, hit.end), quote);
|
|
});
|
|
|
|
test("a bad range is rejected", () => {
|
|
let threw = false;
|
|
try {
|
|
selectionSelectors("abc", 2, 1);
|
|
} catch (e) {
|
|
threw = true;
|
|
eq(e.message, "bad range");
|
|
}
|
|
eq(threw, true);
|
|
});
|