Initial commit

node_modules/execa/lib/io/iterate.js

@@ -0,0 +1,110 @@
+import {on} from 'node:events';
+import {getDefaultHighWaterMark} from 'node:stream';
+import {getEncodingTransformGenerator} from '../transform/encoding-transform.js';
+import {getSplitLinesGenerator} from '../transform/split.js';
+import {transformChunkSync, finalChunksSync} from '../transform/run-sync.js';
+
+// Iterate over lines of `subprocess.stdout`, used by `subprocess.readable|duplex|iterable()`
+export const iterateOnSubprocessStream = ({subprocessStdout, subprocess, binary, shouldEncode, encoding, preserveNewlines}) => {
+	const controller = new AbortController();
+	stopReadingOnExit(subprocess, controller);
+	return iterateOnStream({
+		stream: subprocessStdout,
+		controller,
+		binary,
+		shouldEncode: !subprocessStdout.readableObjectMode && shouldEncode,
+		encoding,
+		shouldSplit: !subprocessStdout.readableObjectMode,
+		preserveNewlines,
+	});
+};
+
+const stopReadingOnExit = async (subprocess, controller) => {
+	try {
+		await subprocess;
+	} catch {} finally {
+		controller.abort();
+	}
+};
+
+// Iterate over lines of `subprocess.stdout`, used by `result.stdout` and the `verbose: 'full'` option.
+// Applies the `lines` and `encoding` options.
+export const iterateForResult = ({stream, onStreamEnd, lines, encoding, stripFinalNewline, allMixed}) => {
+	const controller = new AbortController();
+	stopReadingOnStreamEnd(onStreamEnd, controller, stream);
+	const objectMode = stream.readableObjectMode && !allMixed;
+	return iterateOnStream({
+		stream,
+		controller,
+		binary: encoding === 'buffer',
+		shouldEncode: !objectMode,
+		encoding,
+		shouldSplit: !objectMode && lines,
+		preserveNewlines: !stripFinalNewline,
+	});
+};
+
+const stopReadingOnStreamEnd = async (onStreamEnd, controller, stream) => {
+	try {
+		await onStreamEnd;
+	} catch {
+		stream.destroy();
+	} finally {
+		controller.abort();
+	}
+};
+
+const iterateOnStream = ({stream, controller, binary, shouldEncode, encoding, shouldSplit, preserveNewlines}) => {
+	const onStdoutChunk = on(stream, 'data', {
+		signal: controller.signal,
+		highWaterMark: HIGH_WATER_MARK,
+		// Backward compatibility with older name for this option
+		// See https://github.com/nodejs/node/pull/52080#discussion_r1525227861
+		// @todo Remove after removing support for Node 21
+		highWatermark: HIGH_WATER_MARK,
+	});
+	return iterateOnData({
+		onStdoutChunk,
+		controller,
+		binary,
+		shouldEncode,
+		encoding,
+		shouldSplit,
+		preserveNewlines,
+	});
+};
+
+export const DEFAULT_OBJECT_HIGH_WATER_MARK = getDefaultHighWaterMark(true);
+
+// The `highWaterMark` of `events.on()` is measured in number of events, not in bytes.
+// Not knowing the average amount of bytes per `data` event, we use the same heuristic as streams in objectMode, since they have the same issue.
+// Therefore, we use the value of `getDefaultHighWaterMark(true)`.
+// Note: this option does not exist on Node 18, but this is ok since the logic works without it. It just consumes more memory.
+const HIGH_WATER_MARK = DEFAULT_OBJECT_HIGH_WATER_MARK;
+
+const iterateOnData = async function * ({onStdoutChunk, controller, binary, shouldEncode, encoding, shouldSplit, preserveNewlines}) {
+	const generators = getGenerators({
+		binary,
+		shouldEncode,
+		encoding,
+		shouldSplit,
+		preserveNewlines,
+	});
+
+	try {
+		for await (const [chunk] of onStdoutChunk) {
+			yield * transformChunkSync(chunk, generators, 0);
+		}
+	} catch (error) {
+		if (!controller.signal.aborted) {
+			throw error;
+		}
+	} finally {
+		yield * finalChunksSync(generators);
+	}
+};
+
+const getGenerators = ({binary, shouldEncode, encoding, shouldSplit, preserveNewlines}) => [
+	getEncodingTransformGenerator(binary, encoding, !shouldEncode),
+	getSplitLinesGenerator(binary, preserveNewlines, !shouldSplit, {}),
+].filter(Boolean);