/**
 * @copyright Sister Software
 * @license AGPL-3.0
 * @author Teffen Ellis, et al.
 *
 *   Build a parquet slice from the DeepSeek-generated kryptonite JSONL and emit the corpus-v0.4.0
 *   MANIFEST. corpus-v0.4.0 is a pure adapter-addition revision: it points at every slice from
 *   v0.3.0 plus the new kryptonite slice(s). No v0.3.0 bytes are touched or re-shuffled.
 *
 *   See docs/engineering/reference/CORPUS_V0_4_0_GENERATION.md for the why; that doc also pins the
 *   DeepSeek model version + prompt versions used to produce the JSONL.
 *
 *   Invoke via `mailwoman corpus slice kryptonite \
 *   --jsonl /data/corpus/versioned/v0.4.0/kryptonite/canonical-kryptonite.jsonl \
 *   --base-manifest /data/corpus/versioned/v0.3.0/corpus-v0.3.0/MANIFEST.json \
 *   --out-dir /data/corpus/versioned/v0.4.0`
 */

import { pathExists, readLocalJSONFile } from "@mailwoman/core/fs/readers"
import { writeLocalJSONFile, writeLocalTextFile, makeDirectories } from "@mailwoman/core/fs/writers"
import { join } from "path-ts"
import { JSONSpliterator } from "spliterator"

import type { CanonicalRow, LabeledRow } from "#types"
import { alignRow, PARQUET_COLUMNS, ROW_GROUP_SIZE, SLICE_COMPRESSION, writeSlices } from "#utils"
import type { SliceManifest } from "#utils"

export interface SliceKryptoniteOptions {
	jsonl: string
	baseManifest: string
	outDir: string
	/**
	 * Default `"0.4.0"`.
	 */
	corpusVersion?: string
	/**
	 * Default `"deepseek-kryptonite"`.
	 */
	source?: string
}

async function* canonicalRows(jsonl: string, corpusVersion: string): AsyncIterable<CanonicalRow> {
	for await (const raw of JSONSpliterator.fromAsync<Record<string, unknown>>(jsonl)) {
		// Strip sidecar underscore-prefixed fields the generator left behind for debugging.
		const components = raw["components"] as Record<string, string>

		yield {
			raw: raw["raw"] as string,
			components,
			country: (raw["country"] as string) ?? "US",
			locale: (raw["locale"] as string) ?? undefined,
			source: (raw["source"] as string) ?? "deepseek-kryptonite",
			source_id: raw["source_id"] as string,
			corpus_version: corpusVersion,
			license: (raw["license"] as string) ?? "Synthetic (DeepSeek-v4-flash, AGPL-compatible)",
			synth: raw["synth"] as CanonicalRow["synth"],
		}
	}
}

async function* labeledRows(jsonl: string, corpusVersion: string, quarantineLog: string[]): AsyncIterable<LabeledRow> {
	for await (const row of canonicalRows(jsonl, corpusVersion)) {
		const result = alignRow(row)

		if (result.kind === "labeled") {
			yield result.row
		} else {
			quarantineLog.push(`${row.source_id}\t${result.row.reason}`)
		}
	}
}

export async function buildKryptoniteSlice(
	options: SliceKryptoniteOptions,
	report?: (line: string) => void
): Promise<void> {
	const corpusVersion = options.corpusVersion ?? "0.4.0"
	const source = options.source ?? "deepseek-kryptonite"

	if (!(await pathExists(options.jsonl))) throw new Error(`jsonl not found: ${options.jsonl}`)

	if (!(await pathExists(options.baseManifest))) throw new Error(`base-manifest not found: ${options.baseManifest}`)

	await makeDirectories(options.outDir)

	const quarantine: string[] = []

	const newManifest = await writeSlices(
		{ train: labeledRows(options.jsonl, corpusVersion, quarantine) },
		{ outputDir: options.outDir, corpusVersion }
	)

	report?.(
		`wrote ${newManifest.total_rows} rows into ${newManifest.slices.length} slice(s); ` +
			`quarantined ${quarantine.length}`
	)

	if (quarantine.length) {
		const qPath = join(options.outDir, `corpus-v${corpusVersion}`, "quarantine-kryptonite.tsv")
		await writeLocalTextFile(quarantine.join("\n") + "\n", qPath)
		report?.(`quarantine log → ${qPath}`)
	}

	// Stamp the new slice's source field for audit.ts (which prefers slice.source over
	// first_source_id-prefix inference). Without this, deepseek-kryptonite IDs would have
	// to match a prefix in KNOWN_SOURCE_PREFIXES — we add it there too as a belt-and-braces.
	for (const sh of newManifest.slices) {
		sh.source = source
	}

	// Compose the final corpus-v0.4.0 manifest: every slice from base + the new slice(s).
	const base = await readLocalJSONFile<SliceManifest>(options.baseManifest)

	const combined: SliceManifest = {
		corpus_version: corpusVersion,
		schema: PARQUET_COLUMNS,
		rows_per_slice: base.rows_per_slice,
		row_group_size: base.row_group_size ?? ROW_GROUP_SIZE,
		slices: [...base.slices, ...newManifest.slices],
		counts: {
			train: base.counts.train + (newManifest.counts.train ?? 0),
			val: base.counts.val,
			test: base.counts.test,
		},
		total_rows: base.total_rows + newManifest.total_rows,
	}

	// Stamp source on the legacy v0.3.0 slices too, so audit's slice.source path is the
	// authoritative one. v0.3.0 slices mix sources; we use the first_source_id-prefix
	// inference for them (audit.ts will re-derive on its own when slice.source is absent).
	const combinedPath = join(options.outDir, `corpus-v${corpusVersion}`, "MANIFEST.json")
	await writeLocalJSONFile(combined, combinedPath)
	report?.(`wrote combined manifest → ${combinedPath}`)
	report?.(`  total_rows=${combined.total_rows} (base=${base.total_rows}, added=${newManifest.total_rows})`)
	report?.(`  slices=${combined.slices.length} (base=${base.slices.length}, added=${newManifest.slices.length})`)
	report?.(`  compression=${SLICE_COMPRESSION}`)
}