agentfootprint/dist/time/textForms.js

The explainable agent framework — backtrack a wrong answer to the exact context that caused it (evidence, not guesses). Built on footprintjs.

"use strict";
/**
 * time/textForms — the spellings of the times written in TEXT (time
 * design § 9.5, step T7): the text rule of `forms.ts` · `timeFormsOf`, and
 * the small spelling parts both halves share.
 *
 * Pattern: pure regular expressions over the text — no reading, no resolver.
 * Role:    split from `forms.ts` BY FILE for the synchronous doors: the
 *          evidence gate's exempt corpus (`evidence/evidenceIndex.ts` ·
 *          `addExempt`) runs on every agent with a gate, and a bundler places
 *          a whole file on the synchronous graph — so the window half, which
 *          re-resolves a recorded reading (`resolve.ts`), stays behind
 *          `import()` (the optional-family law,
 *          `test/trace-toolpack/browserGraph.test.ts`). Imports nothing.
 * Emits:   N/A.
 *
 * @example
 * ```ts
 * timeFormsOfText('2026-10-09').said; // ['2026', '10', '9']
 * timeFormsOfText('took 2h');         // { said: [], derived: [] }
 * ```
 */
Object.defineProperty(exports, "__esModule", { value: true });
exports.Spellings = exports.pad2 = exports.clockSpellings = exports.timeFormsOfText = void 0;
// ─── The person's text ─────────────────────────────────────────────────────
/** An ISO calendar date in text — `2026-10-09`, also the date half of `2026-10-09T08:00`. */
const ISO_DATE = /(?<![\d-])(\d{4})-(\d{2})-(\d{2})(?![\d])/g;
/**
 * A 12-hour clock reading in text — `8 Am`, `8:40 AM`, `8pm`, `8:40 p.m.`.
 * Read off the TEXT, not tokens, because the suffix is its own token.
 */
const TWELVE_HOUR = /(?<![\d:.])(\d{1,2})(?::(\d{2}))?\s*([ap])\.?m(?![a-z])/gi;
/**
 * The evidence tokenizer's boundary (`lib/normalize.ts` · `tokenize`),
 * restated for this leaf so a 24-hour reading is recognised as the same
 * whole token the gate compares — pinned equal by `test/time/forms.test.ts`.
 */
const NOT_IN_TOKEN = /[^A-Za-z0-9:_\-/.,%+@#$]+/;
const LIST_COMMA = /,(?!\d)|(?<!\d),/;
const LEADING = /^[$#@+'"`([{<]+/;
/** The trailing decoration `normalize.ts` drops, as character codes. */
const TRAILING = new Set(Array.from('.,;:!?%\'"`)]}>', (char) => char.charCodeAt(0)));
const TWENTY_FOUR_HOUR = /^(\d{1,2}):(\d{2})$/;
/**
 * `piece.replace(/[.,;:!?%'"`)\]}>]+$/, '')`, walked in from the end once —
 * `normalize.ts` · `normalizeToken` drops the same run the same way. The
 * regex retried from every position of a long dot or comma run, which a token
 * can be. It is the walk of `lib/linearText.ts` · `trimTrailing`, restated
 * because this folder imports nothing; its counted work is pinned with the
 * other doors in `test/security/linear-text.test.ts`.
 */
function withoutTrailingDecoration(piece) {
    let end = piece.length;
    while (end > 0 && TRAILING.has(piece.charCodeAt(end - 1)))
        end--;
    return end === piece.length ? piece : piece.slice(0, end);
}
/** The whole tokens of `text` that are a 24-hour clock reading, lower-cased as the gate reads them. */
function clockTokens(text) {
    const out = [];
    for (const rough of text.split(NOT_IN_TOKEN)) {
        for (const piece of rough.split(LIST_COMMA)) {
            const token = withoutTrailingDecoration(piece.toLowerCase().trim().replace(LEADING, ''));
            if (TWENTY_FOUR_HOUR.test(token))
                out.push(token);
        }
    }
    return out;
}
/**
 * The spellings of the times written in `text` — `said` only (the text rule of
 * `forms.ts` · `timeFormsOf`, which delegates here): an ISO date → its year,
 * month and day; a 12-hour or 24-hour clock reading → its spellings. Slash
 * dates and durations are never read.
 *
 * @example
 * ```ts
 * timeFormsOfText('what connected 8 Am to 8:40 AM PST').said;
 * // ['8:00', '08:00', '8:00am', '8:40', '08:40', '8:40am']
 * ```
 */
function timeFormsOfText(text) {
    return { said: textForms(text), derived: [] };
}
exports.timeFormsOfText = timeFormsOfText;
/** The said spellings of every date and clock time written in `text` (the file header's text rule). */
function textForms(text) {
    const out = new Spellings();
    for (const [, year = '', mm, dd] of text.matchAll(ISO_DATE)) {
        const month = Number(mm);
        const day = Number(dd);
        if (month < 1 || month > 12 || day < 1 || day > 31)
            continue;
        out.push(year, String(month), String(day));
    }
    for (const [, hh, mm, half = ''] of text.matchAll(TWELVE_HOUR)) {
        const hour = Number(hh);
        const minutes = mm ?? '00';
        if (hour < 1 || hour > 12 || Number(minutes) > 59)
            continue;
        const h24 = (hour % 12) + (half.toLowerCase() === 'p' ? 12 : 0);
        out.push(...clockSpellings(h24, minutes));
    }
    // A reading the 12-hour pass already took is not read again as 24-hour:
    // `8:40 p.m.` is 20:40, never also 08:40.
    for (const token of clockTokens(text.replace(TWELVE_HOUR, ' '))) {
        const [, hh = '', minutes = ''] = TWENTY_FOUR_HOUR.exec(token) ?? [];
        const h24 = Number(hh);
        if (h24 > 23 || Number(minutes) > 59)
            continue;
        out.push(...clockSpellings(h24, minutes));
    }
    return out.list();
}
/**
 * The spellings of ONE clock time: the 24-hour form bare and padded
 * (`8:00`, `08:00`, `20:00`), the 12-hour colon form (`8:00`) and its glued
 * suffix form (`8:00pm`). After tokenizing, `8:00 PM` and `8:00 AM` both read
 * `8:00` — the same hour on a 12-hour dial.
 */
function clockSpellings(h24, minutes) {
    const h12 = h24 % 12 === 0 ? 12 : h24 % 12;
    const suffix = h24 < 12 ? 'am' : 'pm';
    return [
        `${h24}:${minutes}`,
        `${(0, exports.pad2)(h24)}:${minutes}`,
        `${h12}:${minutes}`,
        `${h12}:${minutes}${suffix}`,
    ];
}
exports.clockSpellings = clockSpellings;
// ─── Small parts ─────────────────────────────────────────────────────────────
const pad2 = (n) => String(n).padStart(2, '0');
exports.pad2 = pad2;
/** An insertion-ordered set of spellings. */
class Spellings {
    seen = new Set();
    push(...forms) {
        for (const form of forms)
            if (form !== '')
                this.seen.add(form);
    }
    list() {
        return [...this.seen];
    }
}
exports.Spellings = Spellings;
//# sourceMappingURL=textForms.js.map