std/tar/tar_stream.ts

570 lines
16 KiB
TypeScript
Raw Normal View History

feat(archive): `UntarStream` and `TarStream` (#4548) * refactor(archive): An implementation of Tar as streams * fmt(archive): Ran `deno fmt` * fix(archive): fixed JSDoc examples in tar_streams.ts * fix(archive): fixed JSDoc examples so `deno task test` doesn't complain * fix(archive): lint license error * fix(archive): lint error files not exported * set(archive): Set current time as mtime for default * resolve(archive): resolves comments made * add(archive): `{ mode: 'byob' }` support for TarStream * add(archive): `{ mode: 'byob' }` support for UnTarStream * adjust(archive): The logical flow of a few if statements * tests(archive): Updated Tests for Un/TarStream * fix(archive): TarStream mtime wasn't an octal * fix(archive): TarStream tests * add(archive): Added parsePathname function Added parsePathname function abstracting the logic out of TarStream allowing the developer to validate pathnames before providing them to TarStream hoping it doesn't throw an error and require the archive creation to start all over again. * fix(archive): extra bytes incorrectly appending at the end of files When the appending file was exactly divisible by 512 bytes, an extra 512 bytes was being appending instead of zero to fill in the gap, causing the next header to be read at the wrong place. * adjust(archive): to always return the amount of bytes requested Instead of using enqueue, the leftover bytes are saved for later for the next buffer provided. * tweaks * fix * docs(archive): Link to the spec that they're following * docs(archive): fix spelling * add(archive): function validTarSteamOptions - To make sure, if TarStreamOptions are being provided, that they are in the correct format so as to not create bad tarballs. * add(archive): more tests * fix(archive): validTarStreamOptions * add(archive): tests for validTarStreamOptions * refactor(archive): code to copy the changes made in the @doctor/tar-stream version * test(archive): added from @doctor/tar-stream * chore: nit on anonymous function * refactor(archive): UnTarStream that fixes unexplainable memory leak - The second newest test introduced here '... with invalid ending' seems to detect a memory leak due to an invalid tarball. I couldn't figure out why the memory leak was happening but I know this restructure of the code doesn't have that same memory leak. * chore: fmt * tests(archive): added remaining tests to cover many lines as possible * adjust(archive): remove simplify pathname code * adjust(archive): remove checking for duplicate pathnames in taring process * adjust(archive): A readable will exist on TarEntry unless string values 1-6 * tests(archive): added more tests for higher coverage * adjust(archives): TarStream and UnTarStream to implement TransformStream * docs(archive): moved TarStreamOptions docs into properties. * adjust(archive): TarStreamFile to take a ReadableSteam instead of an Iterable | AsyncIterable * adjust(archive): to use FixedChunkStream instead of rolling it's own implementation * fix(archive): lint error * adjust(archive): Error types and messages * adjust(archive): more Error messages / improve tests * refactor(archive): UnTarStream to return TarStreamChunk instead of TarStreamEntry * fix(archive): JSDoc example * adjust(archive): mode, uid, gid options to be provided as numbers instead of strings. * adjust(archive): TarStream's pathname to be only of type string * fix(archive): prefix/name to ignore everything past the first NULL * adjust(archive): `checksum` and `pad` to not be exposed from UnTarStream * adjust(archive): checksum calculation * change(archive): `.slice` to `.subarray` * doc(archive): "octal number" to "octal literal" * adjust(archive): TarStreamOptions to be optional with defaults * doc(archive): added more docs for the interfaces * docs(archive): denoting defaults * docs(archive): updated for new lint rules * adjust(archive): Tests to use assertRejects where appropriate & add `validPathname` function - The `validPathname` is meant to be a nicer exposed function for users of this lib to validate that their pathnames are valid before pipping it through the TarStream, over exposing parsePathname where the user may be confused about what to do with the result. * adjust(archive): to use `Date.now()` instead of `new Date().getTime()` Co-authored-by: ud2 <sjx233@qq.com> * adjust(archive): mode, uid, and gid to be numbers instead of strings when Untaring * tests(archive): adjust two tests to also validate the contents of the files are valid * adjust(archive): linkname, uname, and gname to follow the same decoding rules as name and prefix * rename(archive): UnTarStream to UntarStream * fix(archive): type that was missed getting updated * tests(archive): adjust check headers test to validate all header properties instead of just pathnames * rename(archive): `pathname` properties to `path` * docs(archive): updated to be more descriptive * docs(archive): Updated error types * adjust(archive): `validPath` to `assertValidPath` * adjust(archive): `validTarStreamOptions` to `assertValidTarStreamOptions` * revert(archive): UntarStream to produce TarStreamEntry instead of TarStreamChunk * refactor: remove redundant `void` return types * docs: cleanup assertion function docs * docs: correct `TarStream` example * docs: minor docs cleanups * refactor: improve error class specificity * docs: add `@experimental` JSDoc tags * docs(archive): Updated examples for `assertValidPath` and `assertValidTarStreamOptions``` * fix(archive): problem with tests - I suspect the problem is that a file that was read by `Deno.readDir` changed size between being read at `Deno.stat` and when `Deno.open` finished pulling it all in. * update error messages * update error messages * fix typos * refactor: tweak error messages * refactor: tweaks and add type field --------- Co-authored-by: Asher Gomez <ashersaupingomez@gmail.com> Co-authored-by: ud2 <sjx233@qq.com> Co-authored-by: Yoshiya Hinosawa <stibium121@gmail.com>
2024-09-02 07:43:22 +00:00
// Copyright 2018-2024 the Deno authors. All rights reserved. MIT license.
/**
* The interface required to provide a file.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*/
export interface TarStreamFile {
/**
* The type of the input.
*/
type: "file";
/**
* The path to the file, relative to the archive's root directory.
*/
path: string;
/**
* The size of the file in bytes.
*/
size: number;
/**
* The contents of the file.
*/
readable: ReadableStream<Uint8Array>;
/**
* The metadata of the file.
*/
options?: TarStreamOptions;
}
/**
* The interface required to provide a directory.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*/
export interface TarStreamDir {
/**
* The type of the input.
*/
type: "directory";
/**
* The path of the directory, relative to the archive's root directory.
*/
path: string;
/**
* The metadata of the directory.
*/
options?: TarStreamOptions;
}
/**
* A union type merging all the TarStream interfaces that can be piped into the
* TarStream class.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*/
export type TarStreamInput = TarStreamFile | TarStreamDir;
type TarStreamInputInternal =
& (Omit<TarStreamFile, "path"> | Omit<TarStreamDir, "path">)
& { path: [Uint8Array, Uint8Array] };
/**
* The options that can go along with a file or directory.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*/
export interface TarStreamOptions {
/**
* An octal literal.
* Defaults to 755 for directories and 644 for files.
*/
mode?: number;
/**
* An octal literal.
* @default {0}
*/
uid?: number;
/**
* An octal literal.
* @default {0}
*/
gid?: number;
/**
* A number of seconds since the start of epoch. Avoid negative values.
* Defaults to the current time in seconds.
*/
mtime?: number;
/**
* An ASCII string. Should be used in preference of uid.
* @default {''}
*/
uname?: string;
/**
* An ASCII string. Should be used in preference of gid.
* @default {''}
*/
gname?: string;
/**
* The major number for character device.
* @default {''}
*/
devmajor?: string;
/**
* The minor number for block device entry.
* @default {''}
*/
devminor?: string;
}
const SLASH_CODE_POINT = "/".charCodeAt(0);
/**
* ### Overview
* A TransformStream to create a tar archive. Tar archives allow for storing
* multiple files in a single file (called an archive, or sometimes a tarball).
* These archives typically have a single '.tar' extension. This
* implementation follows the [FreeBSD 15.0](https://man.freebsd.org/cgi/man.cgi?query=tar&sektion=5&apropos=0&manpath=FreeBSD+15.0-CURRENT) spec.
*
* ### File Format & Limitations
* The ustar file format is used for creating the tar archive. While this
* format is compatible with most tar readers, the format has several
* limitations, including:
* - Paths must be at most 256 characters.
* - Files must be at most 8 GiBs in size, or 64 GiBs if `sizeExtension` is set
* to true.
* - Sparse files are not supported.
*
* ### Usage
* TarStream may throw an error for several reasons. A few of those are:
* - The path is invalid.
* - The size provided does not match that of the iterable's length.
*
* ### Compression
* Tar archives are not compressed by default. If you'd like to compress the
* archive, you may do so by piping it through a compression stream.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*
* @example Usage
* ```ts ignore
* import { TarStream, type TarStreamInput } from "@std/tar/tar-stream";
feat(archive): `UntarStream` and `TarStream` (#4548) * refactor(archive): An implementation of Tar as streams * fmt(archive): Ran `deno fmt` * fix(archive): fixed JSDoc examples in tar_streams.ts * fix(archive): fixed JSDoc examples so `deno task test` doesn't complain * fix(archive): lint license error * fix(archive): lint error files not exported * set(archive): Set current time as mtime for default * resolve(archive): resolves comments made * add(archive): `{ mode: 'byob' }` support for TarStream * add(archive): `{ mode: 'byob' }` support for UnTarStream * adjust(archive): The logical flow of a few if statements * tests(archive): Updated Tests for Un/TarStream * fix(archive): TarStream mtime wasn't an octal * fix(archive): TarStream tests * add(archive): Added parsePathname function Added parsePathname function abstracting the logic out of TarStream allowing the developer to validate pathnames before providing them to TarStream hoping it doesn't throw an error and require the archive creation to start all over again. * fix(archive): extra bytes incorrectly appending at the end of files When the appending file was exactly divisible by 512 bytes, an extra 512 bytes was being appending instead of zero to fill in the gap, causing the next header to be read at the wrong place. * adjust(archive): to always return the amount of bytes requested Instead of using enqueue, the leftover bytes are saved for later for the next buffer provided. * tweaks * fix * docs(archive): Link to the spec that they're following * docs(archive): fix spelling * add(archive): function validTarSteamOptions - To make sure, if TarStreamOptions are being provided, that they are in the correct format so as to not create bad tarballs. * add(archive): more tests * fix(archive): validTarStreamOptions * add(archive): tests for validTarStreamOptions * refactor(archive): code to copy the changes made in the @doctor/tar-stream version * test(archive): added from @doctor/tar-stream * chore: nit on anonymous function * refactor(archive): UnTarStream that fixes unexplainable memory leak - The second newest test introduced here '... with invalid ending' seems to detect a memory leak due to an invalid tarball. I couldn't figure out why the memory leak was happening but I know this restructure of the code doesn't have that same memory leak. * chore: fmt * tests(archive): added remaining tests to cover many lines as possible * adjust(archive): remove simplify pathname code * adjust(archive): remove checking for duplicate pathnames in taring process * adjust(archive): A readable will exist on TarEntry unless string values 1-6 * tests(archive): added more tests for higher coverage * adjust(archives): TarStream and UnTarStream to implement TransformStream * docs(archive): moved TarStreamOptions docs into properties. * adjust(archive): TarStreamFile to take a ReadableSteam instead of an Iterable | AsyncIterable * adjust(archive): to use FixedChunkStream instead of rolling it's own implementation * fix(archive): lint error * adjust(archive): Error types and messages * adjust(archive): more Error messages / improve tests * refactor(archive): UnTarStream to return TarStreamChunk instead of TarStreamEntry * fix(archive): JSDoc example * adjust(archive): mode, uid, gid options to be provided as numbers instead of strings. * adjust(archive): TarStream's pathname to be only of type string * fix(archive): prefix/name to ignore everything past the first NULL * adjust(archive): `checksum` and `pad` to not be exposed from UnTarStream * adjust(archive): checksum calculation * change(archive): `.slice` to `.subarray` * doc(archive): "octal number" to "octal literal" * adjust(archive): TarStreamOptions to be optional with defaults * doc(archive): added more docs for the interfaces * docs(archive): denoting defaults * docs(archive): updated for new lint rules * adjust(archive): Tests to use assertRejects where appropriate & add `validPathname` function - The `validPathname` is meant to be a nicer exposed function for users of this lib to validate that their pathnames are valid before pipping it through the TarStream, over exposing parsePathname where the user may be confused about what to do with the result. * adjust(archive): to use `Date.now()` instead of `new Date().getTime()` Co-authored-by: ud2 <sjx233@qq.com> * adjust(archive): mode, uid, and gid to be numbers instead of strings when Untaring * tests(archive): adjust two tests to also validate the contents of the files are valid * adjust(archive): linkname, uname, and gname to follow the same decoding rules as name and prefix * rename(archive): UnTarStream to UntarStream * fix(archive): type that was missed getting updated * tests(archive): adjust check headers test to validate all header properties instead of just pathnames * rename(archive): `pathname` properties to `path` * docs(archive): updated to be more descriptive * docs(archive): Updated error types * adjust(archive): `validPath` to `assertValidPath` * adjust(archive): `validTarStreamOptions` to `assertValidTarStreamOptions` * revert(archive): UntarStream to produce TarStreamEntry instead of TarStreamChunk * refactor: remove redundant `void` return types * docs: cleanup assertion function docs * docs: correct `TarStream` example * docs: minor docs cleanups * refactor: improve error class specificity * docs: add `@experimental` JSDoc tags * docs(archive): Updated examples for `assertValidPath` and `assertValidTarStreamOptions``` * fix(archive): problem with tests - I suspect the problem is that a file that was read by `Deno.readDir` changed size between being read at `Deno.stat` and when `Deno.open` finished pulling it all in. * update error messages * update error messages * fix typos * refactor: tweak error messages * refactor: tweaks and add type field --------- Co-authored-by: Asher Gomez <ashersaupingomez@gmail.com> Co-authored-by: ud2 <sjx233@qq.com> Co-authored-by: Yoshiya Hinosawa <stibium121@gmail.com>
2024-09-02 07:43:22 +00:00
*
* await ReadableStream.from<TarStreamInput>([
* {
* type: "directory",
* path: 'potato/'
* },
* {
* type: "file",
* path: 'deno.json',
* size: (await Deno.stat('deno.json')).size,
* readable: (await Deno.open('deno.json')).readable
* },
* {
* type: "file",
* path: '.vscode/settings.json',
* size: (await Deno.stat('.vscode/settings.json')).size,
* readable: (await Deno.open('.vscode/settings.json')).readable
* }
* ])
* .pipeThrough(new TarStream())
* .pipeThrough(new CompressionStream('gzip'))
* .pipeTo((await Deno.create('./out.tar.gz')).writable)
* ```
*/
export class TarStream implements TransformStream<TarStreamInput, Uint8Array> {
#readable: ReadableStream<Uint8Array>;
#writable: WritableStream<TarStreamInput>;
constructor() {
const { readable, writable } = new TransformStream<
TarStreamInput,
TarStreamInputInternal
>({
transform(chunk, controller) {
if (chunk.options) {
try {
assertValidTarStreamOptions(chunk.options);
} catch (e) {
return controller.error(e);
}
}
if (
"size" in chunk &&
(chunk.size < 0 || 8 ** 12 < chunk.size ||
chunk.size.toString() === "NaN")
) {
return controller.error(
new RangeError(
"Cannot add to the tar archive: The size cannot exceed 64 Gibs",
),
);
}
const path = parsePath(chunk.path);
controller.enqueue({ ...chunk, path });
},
});
this.#writable = writable;
const gen = async function* () {
const encoder = new TextEncoder();
for await (const chunk of readable) {
const [prefix, name] = chunk.path;
const typeflag = "size" in chunk ? "0" : "5";
const header = new Uint8Array(512);
const size = "size" in chunk ? chunk.size : 0;
const options: Required<TarStreamOptions> = {
mode: typeflag === "5" ? 755 : 644,
uid: 0,
gid: 0,
mtime: Math.floor(Date.now() / 1000),
uname: "",
gname: "",
devmajor: "",
devminor: "",
...chunk.options,
};
header.set(name); // name
header.set(
encoder.encode(
options.mode.toString().padStart(6, "0") + " \0" + // mode
options.uid.toString().padStart(6, "0") + " \0" + //uid
options.gid.toString().padStart(6, "0") + " \0" + // gid
size.toString(8).padStart(size < 8 ** 11 ? 11 : 12, "0") +
(size < 8 ** 11 ? " " : "") + // size
options.mtime.toString(8).padStart(11, "0") + " " + // mtime
" ".repeat(8) + // checksum | To be updated later
typeflag + // typeflag
"\0".repeat(100) + // linkname
"ustar\0" + // magic
"00" + // version
options.uname.padEnd(32, "\0") + // uname
options.gname.padEnd(32, "\0") + // gname
options.devmajor.padStart(8, "\0") + // devmajor
options.devminor.padStart(8, "\0"), // devminor
),
100,
);
header.set(prefix, 345); // prefix
// Update Checksum
header.set(
encoder.encode(
header.reduce((x, y) => x + y).toString(8).padStart(6, "0") + "\0",
),
148,
);
yield header;
if ("size" in chunk) {
let size = 0;
for await (const value of chunk.readable) {
size += value.length;
yield value;
}
if (chunk.size !== size) {
throw new RangeError(
`Cannot add to the tar archive: The provided size (${chunk.size}) did not match bytes read from provided readable (${size})`,
);
}
if (chunk.size % 512) {
yield new Uint8Array(512 - size % 512);
}
}
}
yield new Uint8Array(1024);
}();
this.#readable = new ReadableStream({
type: "bytes",
async pull(controller) {
const { done, value } = await gen.next();
if (done) {
controller.close();
return controller.byobRequest?.respond(0);
}
if (controller.byobRequest?.view) {
const buffer = new Uint8Array(controller.byobRequest.view.buffer);
const size = buffer.length;
if (size < value.length) {
buffer.set(value.slice(0, size));
controller.byobRequest.respond(size);
controller.enqueue(value.slice(size));
} else {
buffer.set(value);
controller.byobRequest.respond(value.length);
}
} else {
controller.enqueue(value);
}
},
});
}
/**
* The ReadableStream
*
* @return ReadableStream<Uint8Array>
*
* @example Usage
* ```ts ignore
* import { TarStream } from "@std/tar/tar-stream";
feat(archive): `UntarStream` and `TarStream` (#4548) * refactor(archive): An implementation of Tar as streams * fmt(archive): Ran `deno fmt` * fix(archive): fixed JSDoc examples in tar_streams.ts * fix(archive): fixed JSDoc examples so `deno task test` doesn't complain * fix(archive): lint license error * fix(archive): lint error files not exported * set(archive): Set current time as mtime for default * resolve(archive): resolves comments made * add(archive): `{ mode: 'byob' }` support for TarStream * add(archive): `{ mode: 'byob' }` support for UnTarStream * adjust(archive): The logical flow of a few if statements * tests(archive): Updated Tests for Un/TarStream * fix(archive): TarStream mtime wasn't an octal * fix(archive): TarStream tests * add(archive): Added parsePathname function Added parsePathname function abstracting the logic out of TarStream allowing the developer to validate pathnames before providing them to TarStream hoping it doesn't throw an error and require the archive creation to start all over again. * fix(archive): extra bytes incorrectly appending at the end of files When the appending file was exactly divisible by 512 bytes, an extra 512 bytes was being appending instead of zero to fill in the gap, causing the next header to be read at the wrong place. * adjust(archive): to always return the amount of bytes requested Instead of using enqueue, the leftover bytes are saved for later for the next buffer provided. * tweaks * fix * docs(archive): Link to the spec that they're following * docs(archive): fix spelling * add(archive): function validTarSteamOptions - To make sure, if TarStreamOptions are being provided, that they are in the correct format so as to not create bad tarballs. * add(archive): more tests * fix(archive): validTarStreamOptions * add(archive): tests for validTarStreamOptions * refactor(archive): code to copy the changes made in the @doctor/tar-stream version * test(archive): added from @doctor/tar-stream * chore: nit on anonymous function * refactor(archive): UnTarStream that fixes unexplainable memory leak - The second newest test introduced here '... with invalid ending' seems to detect a memory leak due to an invalid tarball. I couldn't figure out why the memory leak was happening but I know this restructure of the code doesn't have that same memory leak. * chore: fmt * tests(archive): added remaining tests to cover many lines as possible * adjust(archive): remove simplify pathname code * adjust(archive): remove checking for duplicate pathnames in taring process * adjust(archive): A readable will exist on TarEntry unless string values 1-6 * tests(archive): added more tests for higher coverage * adjust(archives): TarStream and UnTarStream to implement TransformStream * docs(archive): moved TarStreamOptions docs into properties. * adjust(archive): TarStreamFile to take a ReadableSteam instead of an Iterable | AsyncIterable * adjust(archive): to use FixedChunkStream instead of rolling it's own implementation * fix(archive): lint error * adjust(archive): Error types and messages * adjust(archive): more Error messages / improve tests * refactor(archive): UnTarStream to return TarStreamChunk instead of TarStreamEntry * fix(archive): JSDoc example * adjust(archive): mode, uid, gid options to be provided as numbers instead of strings. * adjust(archive): TarStream's pathname to be only of type string * fix(archive): prefix/name to ignore everything past the first NULL * adjust(archive): `checksum` and `pad` to not be exposed from UnTarStream * adjust(archive): checksum calculation * change(archive): `.slice` to `.subarray` * doc(archive): "octal number" to "octal literal" * adjust(archive): TarStreamOptions to be optional with defaults * doc(archive): added more docs for the interfaces * docs(archive): denoting defaults * docs(archive): updated for new lint rules * adjust(archive): Tests to use assertRejects where appropriate & add `validPathname` function - The `validPathname` is meant to be a nicer exposed function for users of this lib to validate that their pathnames are valid before pipping it through the TarStream, over exposing parsePathname where the user may be confused about what to do with the result. * adjust(archive): to use `Date.now()` instead of `new Date().getTime()` Co-authored-by: ud2 <sjx233@qq.com> * adjust(archive): mode, uid, and gid to be numbers instead of strings when Untaring * tests(archive): adjust two tests to also validate the contents of the files are valid * adjust(archive): linkname, uname, and gname to follow the same decoding rules as name and prefix * rename(archive): UnTarStream to UntarStream * fix(archive): type that was missed getting updated * tests(archive): adjust check headers test to validate all header properties instead of just pathnames * rename(archive): `pathname` properties to `path` * docs(archive): updated to be more descriptive * docs(archive): Updated error types * adjust(archive): `validPath` to `assertValidPath` * adjust(archive): `validTarStreamOptions` to `assertValidTarStreamOptions` * revert(archive): UntarStream to produce TarStreamEntry instead of TarStreamChunk * refactor: remove redundant `void` return types * docs: cleanup assertion function docs * docs: correct `TarStream` example * docs: minor docs cleanups * refactor: improve error class specificity * docs: add `@experimental` JSDoc tags * docs(archive): Updated examples for `assertValidPath` and `assertValidTarStreamOptions``` * fix(archive): problem with tests - I suspect the problem is that a file that was read by `Deno.readDir` changed size between being read at `Deno.stat` and when `Deno.open` finished pulling it all in. * update error messages * update error messages * fix typos * refactor: tweak error messages * refactor: tweaks and add type field --------- Co-authored-by: Asher Gomez <ashersaupingomez@gmail.com> Co-authored-by: ud2 <sjx233@qq.com> Co-authored-by: Yoshiya Hinosawa <stibium121@gmail.com>
2024-09-02 07:43:22 +00:00
*
* await ReadableStream.from([
* {
* type: "directory",
* path: 'potato/'
* },
* {
* type: "file",
* path: 'deno.json',
* size: (await Deno.stat('deno.json')).size,
* readable: (await Deno.open('deno.json')).readable
* },
* {
* type: "file",
* path: '.vscode/settings.json',
* size: (await Deno.stat('.vscode/settings.json')).size,
* readable: (await Deno.open('.vscode/settings.json')).readable
* }
* ])
* .pipeThrough(new TarStream())
* .pipeThrough(new CompressionStream('gzip'))
* .pipeTo((await Deno.create('./out.tar.gz')).writable)
* ```
*/
get readable(): ReadableStream<Uint8Array> {
return this.#readable;
}
/**
* The WritableStream
*
* @return WritableStream<TarStreamInput>
*
* @example Usage
* ```ts ignore
* import { TarStream } from "@std/tar/tar-stream";
feat(archive): `UntarStream` and `TarStream` (#4548) * refactor(archive): An implementation of Tar as streams * fmt(archive): Ran `deno fmt` * fix(archive): fixed JSDoc examples in tar_streams.ts * fix(archive): fixed JSDoc examples so `deno task test` doesn't complain * fix(archive): lint license error * fix(archive): lint error files not exported * set(archive): Set current time as mtime for default * resolve(archive): resolves comments made * add(archive): `{ mode: 'byob' }` support for TarStream * add(archive): `{ mode: 'byob' }` support for UnTarStream * adjust(archive): The logical flow of a few if statements * tests(archive): Updated Tests for Un/TarStream * fix(archive): TarStream mtime wasn't an octal * fix(archive): TarStream tests * add(archive): Added parsePathname function Added parsePathname function abstracting the logic out of TarStream allowing the developer to validate pathnames before providing them to TarStream hoping it doesn't throw an error and require the archive creation to start all over again. * fix(archive): extra bytes incorrectly appending at the end of files When the appending file was exactly divisible by 512 bytes, an extra 512 bytes was being appending instead of zero to fill in the gap, causing the next header to be read at the wrong place. * adjust(archive): to always return the amount of bytes requested Instead of using enqueue, the leftover bytes are saved for later for the next buffer provided. * tweaks * fix * docs(archive): Link to the spec that they're following * docs(archive): fix spelling * add(archive): function validTarSteamOptions - To make sure, if TarStreamOptions are being provided, that they are in the correct format so as to not create bad tarballs. * add(archive): more tests * fix(archive): validTarStreamOptions * add(archive): tests for validTarStreamOptions * refactor(archive): code to copy the changes made in the @doctor/tar-stream version * test(archive): added from @doctor/tar-stream * chore: nit on anonymous function * refactor(archive): UnTarStream that fixes unexplainable memory leak - The second newest test introduced here '... with invalid ending' seems to detect a memory leak due to an invalid tarball. I couldn't figure out why the memory leak was happening but I know this restructure of the code doesn't have that same memory leak. * chore: fmt * tests(archive): added remaining tests to cover many lines as possible * adjust(archive): remove simplify pathname code * adjust(archive): remove checking for duplicate pathnames in taring process * adjust(archive): A readable will exist on TarEntry unless string values 1-6 * tests(archive): added more tests for higher coverage * adjust(archives): TarStream and UnTarStream to implement TransformStream * docs(archive): moved TarStreamOptions docs into properties. * adjust(archive): TarStreamFile to take a ReadableSteam instead of an Iterable | AsyncIterable * adjust(archive): to use FixedChunkStream instead of rolling it's own implementation * fix(archive): lint error * adjust(archive): Error types and messages * adjust(archive): more Error messages / improve tests * refactor(archive): UnTarStream to return TarStreamChunk instead of TarStreamEntry * fix(archive): JSDoc example * adjust(archive): mode, uid, gid options to be provided as numbers instead of strings. * adjust(archive): TarStream's pathname to be only of type string * fix(archive): prefix/name to ignore everything past the first NULL * adjust(archive): `checksum` and `pad` to not be exposed from UnTarStream * adjust(archive): checksum calculation * change(archive): `.slice` to `.subarray` * doc(archive): "octal number" to "octal literal" * adjust(archive): TarStreamOptions to be optional with defaults * doc(archive): added more docs for the interfaces * docs(archive): denoting defaults * docs(archive): updated for new lint rules * adjust(archive): Tests to use assertRejects where appropriate & add `validPathname` function - The `validPathname` is meant to be a nicer exposed function for users of this lib to validate that their pathnames are valid before pipping it through the TarStream, over exposing parsePathname where the user may be confused about what to do with the result. * adjust(archive): to use `Date.now()` instead of `new Date().getTime()` Co-authored-by: ud2 <sjx233@qq.com> * adjust(archive): mode, uid, and gid to be numbers instead of strings when Untaring * tests(archive): adjust two tests to also validate the contents of the files are valid * adjust(archive): linkname, uname, and gname to follow the same decoding rules as name and prefix * rename(archive): UnTarStream to UntarStream * fix(archive): type that was missed getting updated * tests(archive): adjust check headers test to validate all header properties instead of just pathnames * rename(archive): `pathname` properties to `path` * docs(archive): updated to be more descriptive * docs(archive): Updated error types * adjust(archive): `validPath` to `assertValidPath` * adjust(archive): `validTarStreamOptions` to `assertValidTarStreamOptions` * revert(archive): UntarStream to produce TarStreamEntry instead of TarStreamChunk * refactor: remove redundant `void` return types * docs: cleanup assertion function docs * docs: correct `TarStream` example * docs: minor docs cleanups * refactor: improve error class specificity * docs: add `@experimental` JSDoc tags * docs(archive): Updated examples for `assertValidPath` and `assertValidTarStreamOptions``` * fix(archive): problem with tests - I suspect the problem is that a file that was read by `Deno.readDir` changed size between being read at `Deno.stat` and when `Deno.open` finished pulling it all in. * update error messages * update error messages * fix typos * refactor: tweak error messages * refactor: tweaks and add type field --------- Co-authored-by: Asher Gomez <ashersaupingomez@gmail.com> Co-authored-by: ud2 <sjx233@qq.com> Co-authored-by: Yoshiya Hinosawa <stibium121@gmail.com>
2024-09-02 07:43:22 +00:00
*
* await ReadableStream.from([
* {
* type: "directory",
* path: 'potato/'
* },
* {
* type: "file",
* path: 'deno.json',
* size: (await Deno.stat('deno.json')).size,
* readable: (await Deno.open('deno.json')).readable
* },
* {
* type: "file",
* path: '.vscode/settings.json',
* size: (await Deno.stat('.vscode/settings.json')).size,
* readable: (await Deno.open('.vscode/settings.json')).readable
* }
* ])
* .pipeThrough(new TarStream())
* .pipeThrough(new CompressionStream('gzip'))
* .pipeTo((await Deno.create('./out.tar.gz')).writable)
* ```
*/
get writable(): WritableStream<TarStreamInput> {
return this.#writable;
}
}
function parsePath(
path: string,
): [Uint8Array, Uint8Array] {
const name = new TextEncoder().encode(path);
if (name.length <= 100) {
return [new Uint8Array(0), name];
}
if (name.length > 256) {
throw new RangeError(
`Cannot parse the path as the path length cannot exceed 256 bytes: The path length is ${name.length}`,
);
}
// If length of last part is > 100, then there's no possible answer to split the path
let suitableSlashPos = Math.max(0, name.lastIndexOf(SLASH_CODE_POINT)); // always holds position of '/'
if (name.length - suitableSlashPos > 100) {
throw new RangeError(
`Cannot parse the path as the filename cannot exceed 100 bytes: The filename length is ${
name.length - suitableSlashPos
}`,
);
}
for (
let nextPos = suitableSlashPos;
nextPos > 0;
suitableSlashPos = nextPos
) {
// disclaimer: '/' won't appear at pos 0, so nextPos always be > 0 or = -1
nextPos = name.lastIndexOf(SLASH_CODE_POINT, suitableSlashPos - 1);
// disclaimer: since name.length > 100 in this case, if nextPos = -1, name.length - nextPos will also > 100
if (name.length - nextPos > 100) {
break;
}
}
const prefix = name.slice(0, suitableSlashPos);
if (prefix.length > 155) {
throw new TypeError(
"Cannot parse the path as the path needs to be split-able on a forward slash separator into [155, 100] bytes respectively",
);
}
return [prefix, name.slice(suitableSlashPos + 1)];
}
/**
* Asserts that the path provided is valid for a {@linkcode TarStream}.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*
* It provides a means to check that a path is valid before pipping it through
* the `TarStream`, where if invalid will throw an error. Ruining any progress
* made when archiving.
*
* @param path The path as a string
*
* @example Usage
* ```ts no-assert ignore
* import { assertValidPath, TarStream, type TarStreamInput } from "@std/tar";
feat(archive): `UntarStream` and `TarStream` (#4548) * refactor(archive): An implementation of Tar as streams * fmt(archive): Ran `deno fmt` * fix(archive): fixed JSDoc examples in tar_streams.ts * fix(archive): fixed JSDoc examples so `deno task test` doesn't complain * fix(archive): lint license error * fix(archive): lint error files not exported * set(archive): Set current time as mtime for default * resolve(archive): resolves comments made * add(archive): `{ mode: 'byob' }` support for TarStream * add(archive): `{ mode: 'byob' }` support for UnTarStream * adjust(archive): The logical flow of a few if statements * tests(archive): Updated Tests for Un/TarStream * fix(archive): TarStream mtime wasn't an octal * fix(archive): TarStream tests * add(archive): Added parsePathname function Added parsePathname function abstracting the logic out of TarStream allowing the developer to validate pathnames before providing them to TarStream hoping it doesn't throw an error and require the archive creation to start all over again. * fix(archive): extra bytes incorrectly appending at the end of files When the appending file was exactly divisible by 512 bytes, an extra 512 bytes was being appending instead of zero to fill in the gap, causing the next header to be read at the wrong place. * adjust(archive): to always return the amount of bytes requested Instead of using enqueue, the leftover bytes are saved for later for the next buffer provided. * tweaks * fix * docs(archive): Link to the spec that they're following * docs(archive): fix spelling * add(archive): function validTarSteamOptions - To make sure, if TarStreamOptions are being provided, that they are in the correct format so as to not create bad tarballs. * add(archive): more tests * fix(archive): validTarStreamOptions * add(archive): tests for validTarStreamOptions * refactor(archive): code to copy the changes made in the @doctor/tar-stream version * test(archive): added from @doctor/tar-stream * chore: nit on anonymous function * refactor(archive): UnTarStream that fixes unexplainable memory leak - The second newest test introduced here '... with invalid ending' seems to detect a memory leak due to an invalid tarball. I couldn't figure out why the memory leak was happening but I know this restructure of the code doesn't have that same memory leak. * chore: fmt * tests(archive): added remaining tests to cover many lines as possible * adjust(archive): remove simplify pathname code * adjust(archive): remove checking for duplicate pathnames in taring process * adjust(archive): A readable will exist on TarEntry unless string values 1-6 * tests(archive): added more tests for higher coverage * adjust(archives): TarStream and UnTarStream to implement TransformStream * docs(archive): moved TarStreamOptions docs into properties. * adjust(archive): TarStreamFile to take a ReadableSteam instead of an Iterable | AsyncIterable * adjust(archive): to use FixedChunkStream instead of rolling it's own implementation * fix(archive): lint error * adjust(archive): Error types and messages * adjust(archive): more Error messages / improve tests * refactor(archive): UnTarStream to return TarStreamChunk instead of TarStreamEntry * fix(archive): JSDoc example * adjust(archive): mode, uid, gid options to be provided as numbers instead of strings. * adjust(archive): TarStream's pathname to be only of type string * fix(archive): prefix/name to ignore everything past the first NULL * adjust(archive): `checksum` and `pad` to not be exposed from UnTarStream * adjust(archive): checksum calculation * change(archive): `.slice` to `.subarray` * doc(archive): "octal number" to "octal literal" * adjust(archive): TarStreamOptions to be optional with defaults * doc(archive): added more docs for the interfaces * docs(archive): denoting defaults * docs(archive): updated for new lint rules * adjust(archive): Tests to use assertRejects where appropriate & add `validPathname` function - The `validPathname` is meant to be a nicer exposed function for users of this lib to validate that their pathnames are valid before pipping it through the TarStream, over exposing parsePathname where the user may be confused about what to do with the result. * adjust(archive): to use `Date.now()` instead of `new Date().getTime()` Co-authored-by: ud2 <sjx233@qq.com> * adjust(archive): mode, uid, and gid to be numbers instead of strings when Untaring * tests(archive): adjust two tests to also validate the contents of the files are valid * adjust(archive): linkname, uname, and gname to follow the same decoding rules as name and prefix * rename(archive): UnTarStream to UntarStream * fix(archive): type that was missed getting updated * tests(archive): adjust check headers test to validate all header properties instead of just pathnames * rename(archive): `pathname` properties to `path` * docs(archive): updated to be more descriptive * docs(archive): Updated error types * adjust(archive): `validPath` to `assertValidPath` * adjust(archive): `validTarStreamOptions` to `assertValidTarStreamOptions` * revert(archive): UntarStream to produce TarStreamEntry instead of TarStreamChunk * refactor: remove redundant `void` return types * docs: cleanup assertion function docs * docs: correct `TarStream` example * docs: minor docs cleanups * refactor: improve error class specificity * docs: add `@experimental` JSDoc tags * docs(archive): Updated examples for `assertValidPath` and `assertValidTarStreamOptions``` * fix(archive): problem with tests - I suspect the problem is that a file that was read by `Deno.readDir` changed size between being read at `Deno.stat` and when `Deno.open` finished pulling it all in. * update error messages * update error messages * fix typos * refactor: tweak error messages * refactor: tweaks and add type field --------- Co-authored-by: Asher Gomez <ashersaupingomez@gmail.com> Co-authored-by: ud2 <sjx233@qq.com> Co-authored-by: Yoshiya Hinosawa <stibium121@gmail.com>
2024-09-02 07:43:22 +00:00
*
* const paths = (await Array.fromAsync(Deno.readDir("./")))
* .filter(entry => entry.isFile)
* .map((entry) => entry.name)
* // Filter out any paths that are invalid as they are to be placed inside a Tar.
* .filter(path => {
* try {
* assertValidPath(path);
* return true;
* } catch (error) {
* console.error(error);
* return false;
* }
* });
*
* await Deno.mkdir('./out/', { recursive: true })
* await ReadableStream.from(paths)
* .pipeThrough(
* new TransformStream<string, TarStreamInput>({
* async transform(path, controller) {
* controller.enqueue({
* type: "file",
* path,
* size: (await Deno.stat(path)).size,
* readable: (await Deno.open(path)).readable,
* });
* },
* }),
* )
* .pipeThrough(new TarStream())
* .pipeThrough(new CompressionStream('gzip'))
* .pipeTo((await Deno.create('./out/archive.tar.gz')).writable);
* ```
*/
export function assertValidPath(path: string) {
parsePath(path);
}
/**
* Asserts that the options provided are valid for a {@linkcode TarStream}.
*
* @experimental **UNSTABLE**: New API, yet to be vetted.
*
* @param options The TarStreamOptions
*
* @example Usage
* ```ts no-assert ignore
* import { assertValidTarStreamOptions, TarStream, type TarStreamInput } from "@std/tar";
feat(archive): `UntarStream` and `TarStream` (#4548) * refactor(archive): An implementation of Tar as streams * fmt(archive): Ran `deno fmt` * fix(archive): fixed JSDoc examples in tar_streams.ts * fix(archive): fixed JSDoc examples so `deno task test` doesn't complain * fix(archive): lint license error * fix(archive): lint error files not exported * set(archive): Set current time as mtime for default * resolve(archive): resolves comments made * add(archive): `{ mode: 'byob' }` support for TarStream * add(archive): `{ mode: 'byob' }` support for UnTarStream * adjust(archive): The logical flow of a few if statements * tests(archive): Updated Tests for Un/TarStream * fix(archive): TarStream mtime wasn't an octal * fix(archive): TarStream tests * add(archive): Added parsePathname function Added parsePathname function abstracting the logic out of TarStream allowing the developer to validate pathnames before providing them to TarStream hoping it doesn't throw an error and require the archive creation to start all over again. * fix(archive): extra bytes incorrectly appending at the end of files When the appending file was exactly divisible by 512 bytes, an extra 512 bytes was being appending instead of zero to fill in the gap, causing the next header to be read at the wrong place. * adjust(archive): to always return the amount of bytes requested Instead of using enqueue, the leftover bytes are saved for later for the next buffer provided. * tweaks * fix * docs(archive): Link to the spec that they're following * docs(archive): fix spelling * add(archive): function validTarSteamOptions - To make sure, if TarStreamOptions are being provided, that they are in the correct format so as to not create bad tarballs. * add(archive): more tests * fix(archive): validTarStreamOptions * add(archive): tests for validTarStreamOptions * refactor(archive): code to copy the changes made in the @doctor/tar-stream version * test(archive): added from @doctor/tar-stream * chore: nit on anonymous function * refactor(archive): UnTarStream that fixes unexplainable memory leak - The second newest test introduced here '... with invalid ending' seems to detect a memory leak due to an invalid tarball. I couldn't figure out why the memory leak was happening but I know this restructure of the code doesn't have that same memory leak. * chore: fmt * tests(archive): added remaining tests to cover many lines as possible * adjust(archive): remove simplify pathname code * adjust(archive): remove checking for duplicate pathnames in taring process * adjust(archive): A readable will exist on TarEntry unless string values 1-6 * tests(archive): added more tests for higher coverage * adjust(archives): TarStream and UnTarStream to implement TransformStream * docs(archive): moved TarStreamOptions docs into properties. * adjust(archive): TarStreamFile to take a ReadableSteam instead of an Iterable | AsyncIterable * adjust(archive): to use FixedChunkStream instead of rolling it's own implementation * fix(archive): lint error * adjust(archive): Error types and messages * adjust(archive): more Error messages / improve tests * refactor(archive): UnTarStream to return TarStreamChunk instead of TarStreamEntry * fix(archive): JSDoc example * adjust(archive): mode, uid, gid options to be provided as numbers instead of strings. * adjust(archive): TarStream's pathname to be only of type string * fix(archive): prefix/name to ignore everything past the first NULL * adjust(archive): `checksum` and `pad` to not be exposed from UnTarStream * adjust(archive): checksum calculation * change(archive): `.slice` to `.subarray` * doc(archive): "octal number" to "octal literal" * adjust(archive): TarStreamOptions to be optional with defaults * doc(archive): added more docs for the interfaces * docs(archive): denoting defaults * docs(archive): updated for new lint rules * adjust(archive): Tests to use assertRejects where appropriate & add `validPathname` function - The `validPathname` is meant to be a nicer exposed function for users of this lib to validate that their pathnames are valid before pipping it through the TarStream, over exposing parsePathname where the user may be confused about what to do with the result. * adjust(archive): to use `Date.now()` instead of `new Date().getTime()` Co-authored-by: ud2 <sjx233@qq.com> * adjust(archive): mode, uid, and gid to be numbers instead of strings when Untaring * tests(archive): adjust two tests to also validate the contents of the files are valid * adjust(archive): linkname, uname, and gname to follow the same decoding rules as name and prefix * rename(archive): UnTarStream to UntarStream * fix(archive): type that was missed getting updated * tests(archive): adjust check headers test to validate all header properties instead of just pathnames * rename(archive): `pathname` properties to `path` * docs(archive): updated to be more descriptive * docs(archive): Updated error types * adjust(archive): `validPath` to `assertValidPath` * adjust(archive): `validTarStreamOptions` to `assertValidTarStreamOptions` * revert(archive): UntarStream to produce TarStreamEntry instead of TarStreamChunk * refactor: remove redundant `void` return types * docs: cleanup assertion function docs * docs: correct `TarStream` example * docs: minor docs cleanups * refactor: improve error class specificity * docs: add `@experimental` JSDoc tags * docs(archive): Updated examples for `assertValidPath` and `assertValidTarStreamOptions``` * fix(archive): problem with tests - I suspect the problem is that a file that was read by `Deno.readDir` changed size between being read at `Deno.stat` and when `Deno.open` finished pulling it all in. * update error messages * update error messages * fix typos * refactor: tweak error messages * refactor: tweaks and add type field --------- Co-authored-by: Asher Gomez <ashersaupingomez@gmail.com> Co-authored-by: ud2 <sjx233@qq.com> Co-authored-by: Yoshiya Hinosawa <stibium121@gmail.com>
2024-09-02 07:43:22 +00:00
*
* const paths = (await Array.fromAsync(Deno.readDir('./')))
* .filter(entry => entry.isFile)
* .map(entry => entry.name);
*
* await Deno.mkdir('./out/', { recursive: true })
* await ReadableStream.from(paths)
* .pipeThrough(new TransformStream<string, TarStreamInput>({
* async transform(path, controller) {
* const stats = await Deno.stat(path);
* const options = { mtime: stats.mtime?.getTime()! / 1000 };
* try {
* // Filter out any paths that would have an invalid options provided.
* assertValidTarStreamOptions(options);
* controller.enqueue({
* type: "file",
* path,
* size: stats.size,
* readable: (await Deno.open(path)).readable,
* options,
* });
* } catch (error) {
* console.error(error);
* }
* },
* }))
* .pipeThrough(new TarStream())
* .pipeThrough(new CompressionStream('gzip'))
* .pipeTo((await Deno.create('./out/archive.tar.gz')).writable);
* ```
*/
export function assertValidTarStreamOptions(options: TarStreamOptions) {
if (
options.mode &&
(options.mode.toString().length > 6 ||
!/^[0-7]*$/.test(options.mode.toString()))
) throw new TypeError("Cannot add to the tar archive: Invalid Mode provided");
if (
options.uid &&
(options.uid.toString().length > 6 ||
!/^[0-7]*$/.test(options.uid.toString()))
) throw new TypeError("Cannot add to the tar archive: Invalid UID provided");
if (
options.gid &&
(options.gid.toString().length > 6 ||
!/^[0-7]*$/.test(options.gid.toString()))
) throw new TypeError("Cannot add to the tar archive: Invalid GID provided");
if (
options.mtime != undefined &&
(options.mtime.toString(8).length > 11 ||
options.mtime.toString() === "NaN")
) {
throw new TypeError(
"Cannot add to the tar archive: Invalid MTime provided",
);
}
if (
options.uname &&
// deno-lint-ignore no-control-regex
(options.uname.length > 32 - 1 || !/^[\x00-\x7F]*$/.test(options.uname))
) {
throw new TypeError(
"Cannot add to the tar archive: Invalid UName provided",
);
}
if (
options.gname &&
// deno-lint-ignore no-control-regex
(options.gname.length > 32 - 1 || !/^[\x00-\x7F]*$/.test(options.gname))
) {
throw new TypeError(
"Cannot add to the tar archive: Invalid GName provided",
);
}
if (
options.devmajor &&
(options.devmajor.length > 8)
) {
throw new TypeError(
"Cannot add to the tar archive: Invalid DevMajor provided",
);
}
if (
options.devminor &&
(options.devminor.length > 8)
) {
throw new TypeError(
"Cannot add to the tar archive: Invalid DevMinor provided",
);
}
}