MINSTD

Create a readable stream for a linear congruential pseudorandom number generator (LCG) based on Park and Miller.

Usage

var randomStream = require( '@stdlib/random/streams/minstd' );

randomStream( [options] )

Returns a readable stream for a linear congruential pseudorandom number generator (LCG) based on Park and Miller.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

var iStream;
var stream;

function log( chunk, idx ) {
    console.log( chunk.toString() );
    if ( idx === 10 ) {
        stream.destroy();
    }
}

stream = randomStream();
iStream = inspectStream( log );

stream.pipe( iStream );

The function accepts the following options:

  • objectMode: specifies whether a stream should operate in objectMode. Default: false.
  • encoding: specifies how Buffer objects should be decoded to strings. Default: null.
  • highWaterMark: specifies the maximum number of bytes to store in an internal buffer before ceasing to generate additional pseudorandom numbers.
  • sep: separator used to join streamed data. This option is only applicable when a stream is not in objectMode. Default: '\n'.
  • iter: number of iterations.
  • normalized: boolean indicating whether to return pseudorandom numbers on the interval [0,1).
  • seed: pseudorandom number generator seed.
  • state: an Int32Array containing pseudorandom number generator state. If provided, the function ignores the seed option.
  • copy: boolean indicating whether to copy a provided pseudorandom number generator state. Setting this option to false allows sharing state between two or more pseudorandom number generators and/or streams. Setting this option to true ensures that a stream generator has exclusive control over its internal state. Default: true.
  • siter: number of iterations after which to emit the pseudorandom number generator state. This option is useful when wanting to deterministically capture a stream's underlying PRNG state. Default: 1e308.

To set stream options,

var opts = {
    'objectMode': true,
    'encoding': 'utf8',
    'highWaterMark': 64
};

var stream = randomStream( opts );

By default, the function returns a stream which can generate an infinite number of values (i.e., the stream will never end). To limit the number of generated pseudorandom numbers, set the iter option.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

function log( chunk ) {
    console.log( chunk.toString() );
}

var opts = {
    'iter': 10
};

var stream = randomStream( opts );
var iStream = inspectStream( log );

stream.pipe( iStream );

To return pseudorandom numbers on the interval [0,1), set the normalized option.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

function log( chunk ) {
    console.log( chunk.toString() );
}

var opts = {
    'iter': 10,
    'normalized': true
};

var stream = randomStream( opts );
var iStream = inspectStream( log );

stream.pipe( iStream );

By default, when not operating in objectMode, a returned stream delineates generated pseudorandom numbers using a newline character. To specify an alternative separator, set the sep option.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

function log( chunk ) {
    console.log( chunk.toString() );
}

var opts = {
    'iter': 10,
    'sep': ','
};

var stream = randomStream( opts );
var iStream = inspectStream( log );

stream.pipe( iStream );

To seed the underlying pseudorandom number generator, set the seed option.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

function log( v ) {
    console.log( v );
}

var opts = {
    'objectMode': true,
    'iter': 10,
    'seed': 1234
};

var stream = randomStream( opts );

opts = {
    'objectMode': true
};
var iStream = inspectStream( opts, log );

stream.pipe( iStream );

To return a readable stream with an underlying pseudorandom number generator having a specific initial state, set the state option.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

function log( v ) {
    console.log( v );
}

var opts1 = {
    'objectMode': true,
    'iter': 10
};

var stream = randomStream( opts1 );

var opts2 = {
    'objectMode': true
};
var iStream = inspectStream( opts2, log );

// Stream pseudorandom numbers, thus progressing the underlying generator state:
stream.pipe( iStream );

// Create a new PRNG stream initialized to the last state of the previous stream:
var opts3 = {
    'objectMode': true,
    'iter': 10,
    'state': stream.state
};

stream = randomStream( opts3 );
iStream = inspectStream( opts2, log );

// Stream pseudorandom numbers starting from the last state of the previous stream:
stream.pipe( iStream );
stream.seed

The value used to seed the underlying pseudorandom number generator.

var stream = randomStream();

var seed = stream.seed;
// returns <Int32Array>
stream.seedLength

Length of underlying pseudorandom number generator seed.

var stream = randomStream();

var len = stream.seedLength;
// returns <number>
stream.state

Writable property for getting and setting the underlying pseudorandom number generator state.

var stream = randomStream();

var state = stream.state;
// returns <Int32Array>
stream.stateLength

Length of underlying pseudorandom number generator state.

var stream = randomStream();

var len = stream.stateLength;
// returns <number>
stream.byteLength

Size (in bytes) of underlying pseudorandom number generator state.

var stream = randomStream();

var sz = stream.byteLength;
// returns <number>

randomStream.factory( [options] )

Returns a function for creating readable streams which generate pseudorandom numbers via a linear congruential pseudorandom number generator (LCG) based on Park and Miller.

var opts = {
    'objectMode': true,
    'encoding': 'utf8',
    'highWaterMark': 64
};

var createStream = randomStream.factory( opts );

The method accepts the same options as randomStream().


randomStream.objectMode( [options] )

This method is a convenience function to create streams which always operate in objectMode.

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );

function log( v ) {
    console.log( v );
}

var opts = {
    'iter': 10
};
var stream = randomStream.objectMode( opts );

opts = {
    'objectMode': true
};
var iStream = inspectStream( opts, log );

stream.pipe( iStream );

This method accepts the same options as randomStream(); however, the method will always override the objectMode option in options.


Events

In addition to the standard readable stream events, the following events are supported...

'state'

Emitted after internally generating siter pseudorandom numbers.

var opts = {
    'siter': 10 // emit the PRNG state every 10 pseudorandom numbers
};

var stream = randomStream( opts );

stream.on( 'state', onState );

function onState( state ) {
    // Do something with the emitted state, such as save to file...
}

Notes

  • The underlying pseudorandom number generator has a period of approximately 2.1e9 (see Numerical Recipes in C, 2nd Edition, p. 279).
  • An LCG is fast and uses little memory. On the other hand, because the generator is a simple linear congruential generator, the generator has recognized shortcomings. By today's PRNG standards, the generator's period is relatively short. More importantly, the "randomness quality" of the generator's output is lacking. These defects make the generator unsuitable, for example, in Monte Carlo simulations and in cryptographic applications.
  • If PRNG state is "shared" (meaning a state array was provided during stream creation and not copied) and one sets the generator state to a state array having a different length, the underlying PRNG does not update the existing shared state and, instead, points to the newly provided state array. In order to synchronize PRNG output according to the new shared state array, the state array for each relevant PRNG must be explicitly set.
  • If PRNG state is "shared" and one sets the generator state to a state array of the same length, the PRNG state is updated (along with the state of all other PRNGs sharing the PRNG's state array).
  • In order to capture the PRNG state after a specific number of generated pseudorandom numbers, regardless of internal stream buffering, use the siter option in conjunction with a state event listener. Attempting to capture the underlying PRNG state after reading generated numbers is not likely to give expected results, as internal stream buffering will mean more values have been generated than have been read. Thus, the state returned by the state property will likely reflect a future PRNG state from the perspective of downstream consumers.

Examples

var inspectStream = require( '@stdlib/streams/node/inspect-sink' );
var randomStream = require( '@stdlib/random/streams/minstd' );

function log( v ) {
    console.log( v.toString() );
}

var opts = {
    'objectMode': true,
    'iter': 10
};

var stream = randomStream( opts );

opts = {
    'objectMode': true
};
var iStream = inspectStream( opts, log );

stream.pipe( iStream );

CLI

Usage

Usage: random-minstd [options]

Options:

  -h,  --help               Print this message.
  -V,  --version            Print the package version.
       --sep sep            Separator used to join streamed data. Default: '\n'.
  -n,  --iter iterations    Number of pseudorandom numbers.
       --normalized         Generate pseudorandom numbers on the interval [0,1).
       --seed seed          Pseudorandom number generator seed.
       --state filepath     Path to a file containing the pseudorandom number
                            generator state.
       --snapshot filepath  Output file path for saving the pseudorandom number
                            generator state upon exit.

Notes

  • In accordance with POSIX convention, a trailing newline is always appended to generated output prior to exit.
  • Specifying a "snapshot" file path is useful when wanting to resume pseudorandom number generation due to, e.g., a downstream failure in an analysis pipeline. Before exiting, the process will store the pseudorandom number generator state in a file specified according to a provided file path. Upon loading a snapshot (state), the process will generate pseudorandom numbers starting from the loaded state, thus avoiding having to seed and replay an entire analysis.

Examples

$ random-minstd -n 10 --seed 1234

References

  • Park, S. K., and K. W. Miller. 1988. "Random Number Generators: Good Ones Are Hard to Find." Communications of the ACM 31 (10). New York, NY, USA: ACM: 1192–1201. doi:10.1145/63039.63042.
  • Press, William H., Brian P. Flannery, Saul A. Teukolsky, and William T. Vetterling. 1992. Numerical Recipes in C: The Art of Scientific Computing, Second Edition. Cambridge University Press.
Did you find this page helpful?