@arcsine/nodesh
Advanced tools
Node shell is an npm package aimed at providing bash-like operations/simplicity within the node ecosystem. The goal is to make working with files/folders, http requests, and transformations, as as easy as possible. The library is built upon the async ge
Node shell is an npm package aimed at providing bash-like operations/simplicity within the node ecosystem. The goal is to make working with files/folders, http requests, and transformations, as as easy as possible. The library is built upon the async generation constructs within Ecmascript as well as stream constructs within the node ecosystem. This means the performance is iterative and real-time, just in the same way piping works in a Unix shell.
(remote-tokens.js) Example of processing URLs from the input stream
#!/bin/npx @arcsine/nodesh
stdin // Automatically pipe from stdin
.match('URL', 'extract') // Retain only URL patterns and emit as single values
.fetch() // Request each url that comes through
.tokens(/[^A-Za-z0-9_]+/) // Break down returned webpage into tokens
.trim()
.filter(x =>
x.length >= 6 && // Retain words that are 6 chars or more
x.charAt(0) === x.charAt(0).toUpperCase() // And that start with an uppercase letter
)
.stdout; // Pipe the token stream to stdout
As you can see above, the library's aim is to mimic the pattern of command piping, as well as integrate with stdin/stdout seamlessly. With the shebang applied appropriately, this script can be used just like any other cli command.
Example of integrating node scripts within the shell
$ find . -name '*.ts' |\
cat |\
./remote-tokens.js |\
sort -u
This tools is aimed at simple workflows that normally live within the domain of bash scripts. It is meant to be an alternative of staying in bash or jumping over to another language like python. It's aimed at being able to leverage node libraries and utilities while providing a solid set of foundational elements.
The goal of this tool is not to be:
This tool has aspects of all of the above, but it's primary design goal is to focus on providing simplicity in shell-like interaction. To that end, design decisions were made towards simplicity over performance, and towards common patterns versus being completely configurable.
When solving simple problems involving file systems, file contents, and even http requests, the Unix command line is a great place to operate. The command line is a powerful tool with all of the built in functionality, and the simplicity of the Unix philosophy. As bash scripts grow in complexity, maintenance and understanding tend to drop off quickly. When piping a file through 5-10+ commands, following the logic can be challenging.
Usually at this point, is when I would switch over to something like Python given it's "batteries included" mentality, as it's a perfectly fine language in it's own right. That being said, I find it more and more desirable to be able to leverage common tools/libraries from the node ecosystem in these tasks.
The tool revolves around the use of async generators, as denoted by async function *. This allows for the iterative operation, as well as support for asynchronous operations. This means everything within the framework is non-blocking. This means the primary way of using the framework is by accessing your data as an async generator. The library has built in support for converting basic data types into async generators, as well as built-in support for common patterns.
Example of simple async generator
async function * asyncWorker() {
while (true) {
const result = await longOp();
yield result;
}
}
Out of the box, the following types support an .async property that returns an async generator for the supported type. Currently the supported types are:
Set - This will return an async generator over the set contentsMap - This will return an async generator over the map's entries [key, value]NodeJS:ReadStream - This will return a line-oriented async generator over the read streamExample of read stream
const lineGenerator = fs.createReadStream('data.txt').async;
The following primitives also support .async, but will return a generator that only has
a single value, that of the primitive
StringNumberRegExpIn addition to the built-in functionality, a global function of is declared that will allow any value passed in to be converted to an async generator. If the item is iterable or is a stream, it will return the iteration as a generator, otherwise return the value as a single-valued generator.
Example of simple value
const bigIntGen = of(10000n);
...
const numberGen = (100000).async;
const numberGen2 = of(10000);
Within the framework there are some common enough patterns that exposing them globally proves useful.
get of(): typeof RegisterUtil.of;
Will turn any value into a sequence. If the input value is of type:
Iterable - Returns sequence of elementsAsyncIterable - Returns sequence of elementsAsyncGenerator - Returns sequence of elementsReadable/ReadStream - Returns a sequence of lines read from streamof([1,2,3])
.map(x => x ** 2)
// Should be identical
[1,2,3].async
.map(x => x ** 2)
get registerOperator(): (op: Function) => void;
In the process of using the tool, there may be a need for encapsulating common
operations. By default, wrap provides an easy path for re-using functionality,
but it lacks the clarity of intent enjoyed by the built in operators.
(reverse.js)
class Custom {
reverse() {
return this
.collect() // Gather the entire sequence as an array
.map(x => x.reverse()) // Reverse it
.flatten(); // Flatten it back into a single sequence
}
}
registerOperator(Custom);
module global { // Typescript only
interface AsyncGenerator<T = unknown, TReturn = any, TNext = unknown> extends Custom;
}
require('./reverse')
[1,2,3]
.async
.reverse() // Reverse is now available
get argv(): string[];
The cleaned argv parameters for the running script. Starting at index 0,
is the first meaning parameter for the script. This differs from process.argv
by excluding the executable and script name. This is useful as the script may
be invoked in many different ways and the desire is to limit the amount of
guessing needed to handle inputs appropriately.
NOTE: If you are going to use a command line parsing tool, then you would continue to
use process.argv as normal.
(argv[0] ?? 'Enter a file name:'.prompt())
// Pull in name from argv[0] or prompt if missing
.async
.read() // Read file
get stdin(): AsyncGenerator<string>;
Provides direct access to stdin as sequence of lines
stdin // Stream stdin, one line at a time
.map(line => line.split('').reverse().join('')) // Reverse each line
.stdout // Pipe to stdout
get env(): Record<string, string>;
A case insensitive map for accessing environment variables. Like process.env, but
doesn't require knowledge of the case. Useful for simplifying script interactions.
(env.user_name ?? ask('Enter a user name')) // Prompt user name if there
.async // Can call async on sequences and it will return the same value
.map(userName => ... )
range(stop: number, start?: number, step?: number): AsyncGenerator<number, void, unknown>;
Produces a numeric range, between start (1 by default) and stop (inclusive). A step parameter can be defined to specify the distance between iterated numbers.
range(1, 3)
.map(x => x**2)
// sequence of 1, 4, 9
range(10, 1, 2)
// sequence of 1, 3, 5, 7, 9
The entirety of this project centers on the set of available operators. These operators can be broken into the following groups
The core functionality provides some very basic support for sequences
forEach<T>(this: AsyncGenerator<T>, fn: (x: T) => OrProm<any>): Promise<void>;
This operator is a terminal action that receives each element of the sequence in sequence, but returns no value. This function produces a promise that should be waited on to ensure the sequence is exhausted.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.forEach(console.log) // Will output each line
map<T, U>(this: AsyncGenerator<T>, fn: (x: T) => OrProm<U>): AsyncGenerator<U, void, unknown>;
Converts the sequence of data into another, by applying an operation on each element.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.map(line => line.toUpperCase())
// is now a sequence of all uppercase lines
filter<T>(this: AsyncGenerator<T>, pred: (x: T) => OrProm<boolean>): AsyncGenerator<T, void, unknown>;
Determines if items in the sequence are valid or not. Invalid items are discarded, while valid items are retained.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.filter(x => x.length > 10)
// Will retain all lines that are more than 10 characters
flatten<T, U>(this: AsyncGenerator<U[]> | AsyncGenerator<AsyncGenerator<U>>): AsyncGenerator<U>;
Flattens a sequence of arrays, or a sequence of sequences. This allows for operators that return arrays/sequences, to be able to be represented as a single sequence.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.map(line => line.split(/\s+/g)) // Now a string[] sequence
.flatten() // Now a string sequence for each word in the file
flatMap<T, U>(this: AsyncGenerator<T>, fn: (x: T) => Gen<U> | {async: AsyncGenerator<U>;}): AsyncGenerator<U>;
This is a combination of map and flatten as they are common enough in usage to warrant a
combined operator. This will map the the contents of the sequence (which produces an array
or sequence), and producing a flattened output.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.flatMap(line => line.split(/\s+/g)) // Now a word sequence for the file
reduce<T, U>(this: AsyncGenerator<T>, fn: Reducer<U, T> & {init?: () => U;}, acc?: U): AsyncGenerator<U>;
This is the standard reduce operator and behaves similarly as Array.prototype.reduce. This operator
takes in an accumulation function, which allows for computing a single value based on visiting each element
in the sequence. Given that reduce is a comprehensive and produces a singular value, this operation cannot
stream and will block until the stream is exhausted. Normally it is common to understand map and filter as
being implemented by reduce, but in this situation they behave differently.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.flatMap(line => line.split(/\s+/g)) // Now a string sequence for each word in the file
.reduce((acc, token) => {
acc[token] = (acc[token] ?? 0) + 1;
return acc;
}, {}); // Produces a map of words and their respective frequencies within the document
collect<T>(this: AsyncGenerator<T>): AsyncGenerator<T[]>;
Gathers the entire sequence output as a single array. This is useful if you need the entire stream to perform an action.
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.collect() // Now a sequence with a single array (of all the lines)
.map(lines => lines.join('\n'))
// Produces a single string of the whole file
wrap<T, U>(this: AsyncGenerator<T>, fn: (input: AsyncGenerator<T>) => AsyncGenerator<U>): AsyncGenerator<U, void, unknown>;
This is the simplest mechanism for extending the framework as the operator takes in a function that operates on the sequence of data as a whole. It will consume the sequence and produce an entirely new sequence.
async function translate*(lang, gen) {
for await (const line of gen) {
for (const word of line.split(/\s+/g)) {
const translated = await doTranslate(lang, word);
yield translated;
}
}
}
fs.createReadStream('<file>')
.async // Now a line-oriented sequence
.wrap(translate.bind(null, 'fr')); // Produces a sequence of french-translated word
onError<T>(this: AsyncGenerator<T>, alt: AsyncGenerator<T> | (() => AsyncGenerator<T>)): AsyncGenerator<T, void, unknown>;
If an error occurs, use the provided sequence instead
Some of the most common shell operations are iterating through files,
and operating upon those files. To support this, the framework supports
producing files as a sequence of file objects or filenames, given a file
extension or a regex pattern. With Strings and RegExps supporting the
.async property, these are the most common way of finding files.
read(this: AsyncGenerator<string>, type: 'binary'): AsyncGenerator<Buffer>;
read(this: AsyncGenerator<string>, type?: IOType): AsyncGenerator<string>;
read(this: AsyncGenerator<string>, type: 'line' | 'text'): AsyncGenerator<string>;
This operator will treat the inbound string sequence as file names, and will convert the filename (based on IOType)
line (default) - The sequence will produce as series of lines of texttext - The sequence will produce the entire file contents as a single text stringbinary - The sequence will produce a series of Buffer objects'<file>'
.async // Now a sequence of a single value, a file name
.read('binary') // Read as a series of buffers
.reduce((acc, buffer) => {
return acc + buffer.length;
}, 0); // Count number of bytes in file
'<file>'
.async // Now a sequence of a single value, a file name
.read('text') // Read as a single string
.map(text => text.length); // Count number of characters in file
dir(this: AsyncGenerator<string | RegExp>, config: Omit<DirConfig, 'full'> & {full: true;}): AsyncGenerator<ScanEntry>;
dir(this: AsyncGenerator<string | RegExp>, config: DirConfig): AsyncGenerator<string>;
dir(this: AsyncGenerator<string | RegExp>): AsyncGenerator<string>;
dir(this: AsyncGenerator<string | RegExp>, config?: DirConfig): AsyncGenerator<string | ScanEntry>;
dir provides the ability to recursively search for files within a file system. It expects as the
input sequence type:
string which represents a suffix search on file names (e.g. .csv)RegExp which represents a file pattern to search on (e.g. /path\/sub\/.*[.]js/)In addition to the input sequence type, there is an optional config to affect the output.
By default the output of this sequence will be a series of file names, relative to the process.cwd()
that will be eligible for reading or any other file operation.
'.csv'
.async
.dir({ full: true }) // List all '.csv' files, recursively
.forEach(f => {
// Display the filename, and it's modification time
console.log(f.file, f.stats.mtime);
});
Standard operators regarding common patterns for transformations
notEmpty<T>(this: AsyncGenerator<T>): AsyncGenerator<T>;
This is a special type of filter that excludes null, undefined and ''.
Useful for removing empty values.
'<file>'
.async
.read()
.notEmpty() // Return all non-empty lines of the file
tap<T>(this: AsyncGenerator<T>, visit?: (item: T) => OrProm<any>): AsyncGenerator<T>;
tap provides the ability to inspect the sequence without affecting it's production. The function passed in
can produce a promise that will be waited on, if needed.
'.csv'
.async
.dir()
.tap(({stats}) => collectMetrics(stats))
// Stream unchanged, but was able to track file stat information
unique<T>(this: AsyncGenerator<T>, compare?: (a: T, b: T) => OrProm<boolean>): AsyncGenerator<T>;
unique will ensure the output sequence does not have any consecutive duplicates, similar to the unix uniq command.
The uniqueness is only guaranteed linearly, to allow for streaming. Otherwise this would need to wait
for all data before proceeding. You can also specify a custom equality function as needed.
[1, 2, 2, 3, 4, 5, 5, 1, 7]
.async
.unique() // Will produce [1, 2, 3, 4, 5, 1, 7]
// The final 1 repeats as it's not duplicated in sequence
sort<T>(this: AsyncGenerator<T>, compare?: (a: T, b: T) => number): AsyncGenerator<T, any, any>;
sort is a blocking operation as it requires all the data to be able to sort properly. This means it will wait
on the entire sequence before producing new data. The function operates identically to how Array.prototype.sort behaves.
'<file>'
.async
.read() // Now a sequence of lines
.sort() // Sort lines alphabetically
// Now a sequence of sorted lines
batch<T>(this: AsyncGenerator<T>, size: number): AsyncGenerator<T[]>;
Allows for iterative grouping of streamed data, and produces a sequence of arrays. Each array will be batch sized,
except for the final array which will be at most batch size.
'<file>'
.async
.read() // Generator of file lines
.batch(20) // Generator of array of lines, at most 20 items in length
.map(lines => lines.sort()) // Sort each batch
// Generator of sorted list strings
pair<T, U>(this: AsyncGenerator<T>, value: OrCall<OrGen<U>>, mode?: PairMode): AsyncGenerator<[T, U]>;
pair allows for combining two sets of data into a single sequence of pairs.
The second value can either be a single value, which will be added to every item,
or it could be an iterable element that will match with each item as possible. If the second
iterator runs out, the remaining values can be affected by the mode parameter:
'empty' - Fill in with undefined once the second iterator is exhausted. This is default for iterable values.'repeat' - Loop iteration on the secondary iterator. This is default for string values.'exact' - Stop the emitting values once the secondary iterator is exhausted.'.ts'
.async
.dir() // List all '.ts' files
.flatMap(file => file.async
.read() // Read each file as a sequence of lines
.pair(file) // Combine each line with the file name
.map(([a,b]) => [b, a]) // Reverse the order of the columns
)
// Generator of file lines with, file name attached
Support for common textual operations.
As text operators, these only apply to sequences that produce string values.
columns(this: AsyncGenerator<string>, sep?: RegExp | string): AsyncGenerator<string[]>;
columns is similar to the unix awk in that it allows for production of
columns from a single line of text. This is useful for dealing with column
oriented output. The separator defaults to all whitespace but can tailored
as needed by regex or string.
'<file>.tsv' // Tab-separated file
.async
.read() // Read as lines
.columns('\t') // Separate on tabs
// Now an array of tuples (as defined by tabs in the tsv)
columns<V extends readonly string[]>(this: AsyncGenerator<string>, names: V, sep?: RegExp | string): AsyncGenerator<Record<V[number], string>>;
Supports passing in column names to produce objects instead of tuples. These values will be matched with the columns produced by the separator. Any row that is shorter than the names array will have undefined for the associated keys.
'<file>.tsv' // Tab-separated file
.async
.read() // Read as lines
.columns(['Name', 'Age', 'Major'], '\t') // Separate on tabs
// Now an array of objects { Name: string, Age: string, Major: string } (as defined by tabs in the tsv)
tokens(this: AsyncGenerator<string>, sep?: RegExp | string): AsyncGenerator<string, any, any>;
This operator allows for producing a single sequence of tokens out of lines of text. The default separator is whitespace.
'<file>'
.async
.read() // Read file as lines
.tokens() // Convert to words
.filter(x => x.length > 5) // Retain only words 6-chars or longer
match(this: AsyncGenerator<string>, regex: RegExp | string | RegexType, mode?: MatchMode): AsyncGenerator<string>;
match is similar to tokens, but will emit based on a pattern instead of
just word boundaries.
In addition to simple regex or string patterns, there is built in support for some common use cases (RegexType)
'URL' - Will match on all URLs'EMAIL' - Will match on all emailsAdditionally, mode will determine what is emitted when a match is found (within a single line):
undefined - (default) Return entire line'extract' - Return only matched element'negate' - Return only lines that do not match'<file>'
.async
.read()
.match(/(FIXME|TODO)/, 'negate')
// Exclude all lines that include FIXME or TODO
'<file>'
.async
.read()
.match(/\d{3}(-)?\d{3}(-)?\d{4}/, 'extract)
// Return all phone numbers in the sequence
replace(this: AsyncGenerator<string>, pattern: RegExp | string, sub: string | Replacer): AsyncGenerator<string>;
replace behaves identically to String.prototype.replace, but will only operate
on a single sequence value at a time.
'<file>'
.async
.read()
.replace(/TODO/, 'FIXME')
// All occurrences replaced
trim(this: AsyncGenerator<string>): AsyncGenerator<string>;
trim behaves identically to String.prototype.trim, but will only operate on a single sequence value at a time
'<file>'
.async
.read()
.trim()
// Cleans leading/trailing whitespace per line
singleLine(this: AsyncGenerator<string>): AsyncGenerator<string>;
singleLine is a convenience method for converting an entire block of
text into a single line. This is useful when looking for patterns that
may span multiple lines.
'<file>.html'
.async
.read()
.singleLine() // Convert to a single line
.replace(/<[^>]+?>/) // Remove all HTML tags
join(this: AsyncGenerator<string>, joiner?: string | ((a: string[]) => string)): AsyncGenerator<string>;
This operator allows for combining a sequence of strings into a single value similar to String.prototype.join.
'<file>'
.async
.read() // Read as a series of lines
.join('\n')
// Produces a single value of the entire file
Support for limiting sequence values based on ordering
first<T>(this: AsyncGenerator<T>, n?: number): AsyncGenerator<T, void, unknown>;
This will return the first n elements with a default of a single element.
'<file>'
.async
.read()
.first(10) // Read first 10 lines
'<file>'
.async
.read()
.first() // Read first line
skip<T>(this: AsyncGenerator<T>, n: number): AsyncGenerator<T>;
This will return all but the first n elements.
'<file>.csv'
.async
.read()
.skip(1) // Skip header
last<T>(this: AsyncGenerator<T, any, any>, n?: number): AsyncGenerator<T, any, any>;
This will return the last n elements with a default of a single element.
Since this method requires knowledge of the length of the sequence to
work properly, this now becomes a blocking operator.
'<file>'
.async
.read()
.last(7) // Read last 7 lines of file
'<file>'
.async
.read()
.last() // Read last line of file
repeat<T>(this: AsyncGenerator<T>, n?: number): AsyncGenerator<T>;
This will repeat the first n elements with a default of all elements.
'<file>'
.async
.read()
.first(10) // Read first 10 lines
Support for dealing with execution of external programs
execEach<T>(this: AsyncGenerator<T>, cmd: string, args: string[]): AsyncGenerator<string>;
Execute the command against each item in the sequence. Allow for a list of args to prepend to the command execution. The command's stdout is returned as individual lines.
'.ts'
.async
.dir() // Get all files
.execEach('wc', '-l') // Execute word count for each line
exec(cmd: string, args: string[], output: 'line' | 'text'): AsyncGenerator<string>;
exec(cmd: string, args: string[], output: 'binary'): AsyncGenerator<Buffer>;
exec(cmd: string, args: string[]): AsyncGenerator<string>;
exec(cmd: string): AsyncGenerator<string>;
Pipe the entire sequence as input into the command to be executed. Allow for args to be prepended to the command as needed. If the output is specified as 'binary', the generator will return a sequence of Buffers, otherwise will return strings
'.ts'
.async
.dir() // Get all files
.read() // Read all files
.exec('wc', ['-l']) // Execute word count for all files
// Run in a single operation
Support for exporting data from a sequence
stream<T>(this: AsyncGenerator<T>, mode?: IOType): Readable;
Converts a sequence into a node stream. This readable stream should be
considered standard, and usable in any place a stream is expected. The mode
determines if the stream is string or Buffer oriented.
const stream = '<file>.png'
.async
.read('binary') // Read file as binary
.exec('convert', ['-size=100x20']) // Pipe to convert function
.stream('binary') // Read converted output into NodeJS stream
stream.pipe(fs.createWriteStream('out.png')); // Write out
write<T extends string | Buffer | any>(this: AsyncGenerator<T>, writable: OrStr<Writable>): Writable;
Emits the sequence contents to a write stream. If the write stream is a string, it is considered to be a file name. Buffer contents are written as is. String contents are written as lines.
'<file>.png'
.async
.read('binary') // Read file as binary
.exec('convert', ['-size=100x20']) // Pipe to convert function
.write('out.png') // Write file out
writeFinal(this: AsyncGenerator<string>, file: string): Promise<void>;
export declare class ExportPropOperators<T> {
Writes the entire stream to a file, as a final step. The write stream will not be created until all the values have been emitted. This is useful for reading and writing the same file.
'<file>'
.async
.read()
.replace(/TEMP/, 'final')
.writeFinal('<file>');
get values(this: AsyncGenerator<T>): Promise<T[]>;
Extract all sequence contents into a single array and return as a promise
const values = await '<file>.csv'
.async
.read()
.csv('Width', 'Depth', 'Height'])// Convert to objects
.map(({Width, Height, Depth}) =>
int(Width) * int(Height) * int(Depth) // Compute volume
)
.values // Get all values;
get value(this: AsyncGenerator<T>): Promise<T>;
Extract first sequence element and return as a promise
const name = await 'What is your name?'
.async
.prompt() // Prompt for name
.value // Get single value
get stdout(this: AsyncGenerator<T>): Writable;
Simple method that allows any sequence to be automatically written to stdout
'<file>'
.async
.read() // Read file
.map(line => line.length) // Convert each line to it's length
.stdout // Pipe to stdout
get console(this: AsyncGenerator<T>): Promise<void>;
Simple property that allows any sequence to be automatically called with console.log
'<file>'
.async
.read() // Read file
.json()
.console // Log out objects
FAQs
A node-based library, providing Unix shell-like functionality
The npm package @arcsine/nodesh receives a total of 1 weekly downloads. As such, @arcsine/nodesh popularity was classified as not popular.
We found that @arcsine/nodesh demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Newly introduced telemetry in devenv 1.4 sparked a backlash over privacy concerns, leading to the removal of its AI-powered feature after strong community pushback.
Security News
TC39 met in Seattle and advanced 9 JavaScript proposals, including three to Stage 4, introducing new features and enhancements for a future ECMAScript release.
Security News
Deno 2.2 enhances Node.js compatibility, improves dependency management, adds OpenTelemetry support, and expands linting and task automation for developers.