poteto

poteto allows fetch to work with local files over non-standartized file: protocol.

It can work as fetch polyfill (setting Proxy over globalThis.fetch and globalThis.Request) or as ponyfill (providing a separate drop-in replacements for them).

It can understand:

• URL instance having .protocol === 'file:'
• string or any other object with a stringifier, such as:
  • absolute URL having file protocol, e.g. file:///usr/bin/node
  • relative URL having file protocol, e.g. file:file.ext, file:../subdir/file.ext ¹
  • relative URL, including:
    • relative URL starting from directory or file name, e.g. file.ext or subdir/file.ext
    • relative URL starting from ., e.g. ./file.ext or ../subdir/file.ext
    • URL with absolute pathname starting from /, e.g. /usr/bin/node
    • schemeless URL starting from ///, e.g. ///usr/bin/node
    • empty string, which is equivalent to file: or .

...and work with them in somewhat HTTP-compatible manner.

Relative paths are resolved against CWD².

URLs that are normally supported by standard fetch are fetched like usual.

Zero dependencies. Can work with Node.js's builtin fetch or with user-provided implementation.

Installation

1$ npm i poteto

Usage

1import poteto from 'poteto';
2
3// read files with GET
4{
5  const text = await poteto(import.meta.url).then(r => r.text());
6  const json = await poteto('node_modules/poteto/package.json').then(r => r.json());
7  const buffer = await poteto('/usr/bin/node').then(r => r.arrayBuffer());
8}
9
10// read files with READ in memory-efficient stream-based way
11{
12  const { body } = await poteto('/dev/urandom', { method: 'READ' });
13}
14
15// read file ranges and validate result using SRI
16{
17  const text = await poteto('./node_modules/poteto/LICENSE', {
18    headers: {
19      'Range': 'bytes=34264-34266,17303-17304,19991-19991',
20    },
21    integrity: 'sha512-pmndJoYi4kwRRbKcFCVXRnwT0nF/4d73zYKnqredfUIhfKddjRqL3Zbi+CjjkyMIX2e2HQEkV89kZeWdhj1MJQ==',
22  }).then(r => r.text());
23}
24
25// write files with PUT
26{
27  const { status } = await poteto('file:./new_file.txt', {
28    method: 'PUT',
29    body: 'new ifel contents',
30  });
31}
32
33// write partial file range with POST
34{
35  const { statusText } = await poteto('new_file.txt', {
36    method: 'POST',
37    body: 'fileUNUSED',
38    headers: {
39      'Range': 'bytes=4-7',
40    },
41  });
42}
43
44// write files with WRITE for no particular reason
45{
46  const potentialErrorMessage = await poteto('./new_file.txt', {
47    method: 'WRITE',
48    body: 'newer file contents',
49  }).then(r => r.text());
50}
51
52// append to files with APPEND
53{
54  const { ok } = await poteto('oops/../new_file.txt', {
55    method: 'APPEND',
56    body: '\nbut wait, there is more\n',
57  });
58}
59
60// list files and directories with LIST
61{
62  const filenames = await poteto('.', { method: 'LIST' }).then(r => r.json());
63}
64
65// get stats without reading the file with HEAD
66// (GET, READ and LIST also return fs stats)
67{
68  const { headers } = await poteto('file:package.json', { method: 'HEAD' });
69  const filesize = headers.get('Content-Length');
70  const mtimeHTTPdate = headers.get('Last-Modified');
71  const mtimeTemporal = new Temporal.Instant(headers.get('X-Poteto-MtimeNs'));
72}
73
74// make conditional request (also supports ETag and If-None-Match)
75{
76  const response = await poteto('file:///var/log/auth.log', {
77    headers: {
78      'If-Modified-Since': 'Wed, 07 Aug 2019 10:00:12 GMT',
79    },
80  });
81  const isNotModified = response.status === 304;
82}
83
84// get symlink destination using manual redirect
85{
86  const { headers } = await poteto('/etc/mtab', { redirect: 'manual' });
87  const location = headers.get('Location');
88}
89
90// get errors as JSON
91{
92  const errorInfo = await poteto('file:///non/existent/path', {
93    method: 'PUT',
94    body: 'some data',
95    headers: {
96      'Accept': 'application/json',
97    },
98  }).then(r => r.json());
99}
100
101// and do usual fetches transparently
102{
103  const response = await poteto('https://github.com/LiviaMedeiros/poteto', {
104    headers: {
105      'User-Agent': 'Poteto',
106    },
107  });
108}

Methods

HTTP-alike methods

GET

Reads file. Supports If-Modified-Since, If-Unmodified-Since, If-None-Match, If-Match, Range, and SRI³.

Returns HTTP 200 or HTTP 206 and file body; or HTTP 304.

HEAD

Returns fs.Stats of file in headers. Note that GET also does it.

Returns HTTP 200 without body.

DELETE

Deletes file or directory.

Returns HTTP 204.

POST and PUT

Writes request body to file. Supports Range header (only one range is allowed).

POST opens file with r+ flag while PUT uses w+. Which means:

• POST can write partial file range in existing file without rewriting, but PUT will truncate the end and fill beginning with NUL bytes.
• PUT will create new files or overwrite existing ones, but POST will return HTTP 404 if the file doesn't exist yet.

Returns HTTP 201.

POTETO methods

These are case-sensitive.

APPEND

Appends request body to file.

Returns HTTP 201.

LIST

Reads directory contents. Also gets fs.Stats for the directory.

Returns HTTP 200 with JSON response, containing array of items. Subdirectory names will have trailing /.

READ

Reads file. See Why there is GET and READ.

Returns HTTP 200 and file body.

WRITE

Alias to PUT.

Returns HTTP 201.

Response headers

The following HTTP headers are supported:

• Server with fixed value poteto
• Accept-Ranges with fixed value bytes
• Date with RFC 7231 date of forming the response headers
• Content-Length with filesize in bytes if applicable
• Last-Modified with RFC 7231 date of last file modification if applicable
• ETag with weak entity tag if applicable ⁴

Poteto-specific headers are be prefixed with X-Poteto- prefix. For example, filesize can be accessed via X-Poteto-Size header.

Error handling

If there is an error and it's known, returned promise is resolved with Response having reasonable HTTP status. For example, ENOENT translates to HTTP 404 and EACCES translates to HTTP 403.

The headers contain error information. If Accept header is application/json, the body will contain serialized error as JSON; otherwise, it will contain its .message.

If the error is unknown, returned promise is rejected with it.

Redirect handling

If request.redirect === 'follow' (default), symlinks will be quietly resolved.

If request.redirect === 'error', returned promise will be rejected with TypeError.

If request.redirect === 'manual', returned promise will be resolved with HTTP 302 response with target URL in Location header.

Subpaths

Subpath imports allow to add poteto to the project in different ways.

poteto/polyfill

1import 'poteto/polyfill';
2
3await fetch('/dev/null');
4(new Request('/dev/null')).url === 'file:///dev/null'

Replaces globalThis.fetch and globalThis.Request with proxies. Constructing Request with applicable URL results in regular Request instance but with fully-resolved file: URL. For non-file: URLs, it works transparently.

poteto/ponyfill

1import ponyfill from 'poteto/ponyfill';
2import { fetch as potetoFetch, Request as PotetoRequest } from 'poteto/ponyfill';
3
4ponyfill === potetoFetch;

Provides separate AsyncFunction fetch and Class Request that use globalThis.fetch and globalThis.Request internally but do not allow to indirectly mutate them.

poteto/factory

1import { proxify, wrap } from 'poteto/factory';
2
3const mutableFetch = proxify(fetch);
4const independentFetch = wrap(fetch);

proxify allows to make poteto-like proxy around user-specified fetch function. This will use proxified Request internally. wrap returns a new function that will use user-specified fetch function as fallback for non-file: URLs. This will use PotetoRequest class that extends Request.

poteto/internal

1import { methods, execute } from 'poteto/internal';
2
3methods.set('HELLO', async (url, request) => new Response('WORLD'));
4
5await execute(new Request('file:///dev/zero', { method: 'HELLO' }));

Exposes internal Map methods and AsyncFunction execute.

methods has methods in a form of functions that take the URL instance as first argument and Request instance as second. This was made for convenience, because request.url exposes string instead of URL instance.

execute(request[, url]) performs fetch using the methods map. Unlike in other imports, this one does not fallback for non-file: URL and executes request anyway. Optional url parameter allows to override first argument passing to the method.

Rationale

Sometimes we just want to be able to read local and remote files in the same manner. Some other languages have similar capabilities, for example, file_get_contents() in PHP can read files over http(s).

Sometimes we want to avoid explicitly using node:fs for trivial reading operation.

Sometimes we want convenience methods such as response.json()

Sometimes we don't want to switch context of thinking between node:fs and Fetch API.

Sometimes we have urlstring pointing on file instead of pathstring or URL instance.

Sometimes we have relative URL rather than relative path.

Sometimes we want to read a file requested by URL and have appropriate HTTP Response to return.

HTTP-alike file:

Right now file protocol is not standartized, and HTTP entities such as status codes, headers and algorithms are not related to it.

However, some projections such as GET => read file, ENOENT => 404 Not Found, Accept: application/json => return data as json are intuitive enough to be implemented, so here we are.

CLI tools

There are a few things in /bin/ that work as minimalistic examples.

poteto-cat fileurl1[, fileurl2[, ...]]

Works as cat(1): reads files in order, concatenates and prints on the stdout.

Uses GET method (hence, can be used with https: URLs as is).

poteto-dog fileurl1[, fileurl2[, ...]]

Like poteto-cat, but insane: reads everything in async, and prints chunks on the stdout as fast as it can. If there are multiple files, depending on I/O, they may partially diffuse.

Uses READ method (depending on web server, can be used with https: URLs, but not recommended).

poteto-ls [dirurl1[, dirurl2[, ...]]]

Works like recursive ls(1): reads directories recursively, and outputs as pretty-printed JSON.

1// poteto-ls lib testdir
2{
3 "lib/": [
4  "constants.mjs",
5  "fs.mjs",
6  "generic.mjs",
7  "http.mjs",
8  "methods.mjs",
9  "poteto.mjs",
10  "request.mjs",
11  "sri.mjs"
12 ],
13 "testdir/": [
14  ".keep",
15  {
16   "bin/": [
17    ".keep"
18   ]
19  },
20  {
21   "redirect/": [
22    ".keep",
23    "link1",
24    "link2",
25    "target"
26   ]
27  },
28  {
29   "sequence/": [
30    ".keep"
31   ]
32  }
33 ]
34}

Uses LIST method, don't use on https: URLs.

poteto-put fileurl

Reads data from stdin and prints to file. Overwrites existing files, can create new files, can read from interactive (keyboard) input.

Uses PUT method, can be used with https: URL.

poteto-rm fileurl1[, fileurl2[, ...]]

Works like rm(): deletes files. Not recursive, not interactive.

Uses DELETE method, can be used with https: URLs.

FAQ

Why there is GET and READ

GET loads the whole file in memory before responding. This might be significantly faster, but also means that getting 1Gb file will require >1Gb of memory to be used, no matter what.

Also GET supports Range header and SRI³, and might support other features that require checking file body (e.g. magic to determine Content-Type).

READ is stream-based. If the file body might not be consumed, or consumed partially, or consumed chunk-by-chunk, it will not allocate unnecessarily big amounts of memory.

Also READ leaks filehandles.

Why there is PUT and WRITE

Just to mirror GET and READ. PUT is preferable.

Why is it called like that

No particular reason.

Can this be used in production

Why not, but be aware that there is no standard at the moment.

License

GPL-3.0-or-later