Iter8 NPM | npm.io

iter8

Use iterables, iterators and generators as easily as arrays. Provides the complete Array API (plus a lot more), and works with everything you throw at it.

iter8 is a small (under 4k gzipped) data transformation library that works with JavaScript iterables. It provides the familiar Array API plus a full complement of sequence and set operations like intersect, union, except, and leftJoin.

Features

Works with all iterators, iterables, and generators
Implements complete Array API, plus many additions
Painless interop with Array, Set, and Map types
Deferred execution of all sequence enumeration
Correct return() handling for incomplete iterators
Caches iterator output: automatically re-use sequences without access to the creator
Powerful reflection capability for iterating object properties and getters/setters
Argument type checking
Includes Typescript typings for complete GUI support of its API

Installation

iter8 is on npm:

npm install --save iter8

Standalone packages for use in a browser that create a global iter can be downloaded from github under dist.

Usage

Wrap any iterable object with iter(..) to create an Iter instance, and then use familiar Array methods for transformation. Many new transformation and traversal methods are also added, like groupBy and flatten. All evaluation is performed only when actual data is exposed via methods that return data, and it's all performed via iteration, so only elements actually used by an operation are traversed.

// someArray = [{active, id}, ... ]

const lookup = iter(someArray)
    .filter(e=>e.active)
    .map(e=>[e.id, e])
    .toArray();

const firstThree = lookup.take(3).toArray()  // array of only first 3 elements
const fourth = lookup.get(4)                 // value of 4th element in lookup

Let's do some more interesting things:

// money = [{ category, amount }, ... ]
// group all sale transactions by category, and return the
// # and total $ of sales per category

let money = iter(allSales)
    .groupBy('category')
    .map(([category, sales])=> {
        return {
            volume: sales.length,
            totalDollars: iter(sales).map(e=>e.amount).sum()    
        }
    })

Use data structures to help you:

let money = iter(allSales)
    .groupBy('category')
    .as(Map)

// `money` is a `Map` object with the category as the key 
let foodSales = money.get('food');

API

static methods

These are used to create Iter instances.

instance methods

These methods are available on all Iter instances. Methods identified with an asterisk (*) are essentially the same as their Array prototype equivalent, though there may be minor differences associated with dealing with sequences instead of arrays (see detail).

Iter8 objects have two types of methods: transformation and value-producing.

transformation methods return another Iter instance and can be chained.

value-producing methods return a value directly, and therefore are not chainable.

transformation methods

Mapping/Filtering/Grouping

Merging/Set Operations

Ordering

Traversal

Special

value-producing methods

Element selection

Aggregation/Analysis

Comparison

sequenceEqual(sequence)

Export

Special

forEach(callback, thisArg)*

Usage Details

Execution of every query is deferred until a value producing method is called, which exports data outside the construct of an Iter object. If your query doesn't need to iterate over the entire sequence, it won't. Each value-producing methods returns something other than an Iter object thus ending the chain and causing execution.

This is extremely powerful, but can have unitended side effects, particular when chaining queries with recursion or loops. If you use data passed by reference during a transformation, the value used will be the value at the time the transformation happens -- not when the query was created. You can use the execute method as needed to fully iterate a sequence.

In addition to its own API, Iter implements method for all the Array prototype methods that don't mutate the array. Some of these are value-producing, such as indexOf, while some are transformation functions and produce a new sequence, such as filter.

"key" argument

For many methods that require a key for equality comparison, like groupBy, leftJoin, and set operations, there will be an parameter of type key. This means that the argument can be one of three things:

a function(item) that, given an item in a sequence, returns a key that identifies it
a string, which identifies a property or index on the item in the sequence whose value is the key
null or undefined (or, if the paramter is omitted), meaning the item itself is the key

This means you can unversally use strings to refer to an object property in these situations, so the following are identical:

let x = item.groupBy('name')
let x = item.groupBy(e => e.name)

Note about key-value pairs

When describing an element as a key-value pair, this always means an array with two elements: [key, value]. This is the data structure used by JavaScript Map objects, and is also used for many other operations by Iter such as iterating over objects property, value, and grouping [group-key, group-members].

Note about use of undefined

JavaScript arrays can be sparse, meaning certain indices are not defined. The can also contain undefined elements which are treated differently by some native operations. With sequences, there is no notion of sparseness, and undefined is a perfectly valid value.

Some methods like first could have no value to return, e.g. when called on a zero-length sequence. In JavaScript tradition, iter8 doesn't throw too many errors, but rather returns undefined in situations like this. This can result in indeterminate situations... was there no element, or was the result undefined?

To resolve this, methods like first include an argument that allows you to provide a default value to use other than undefined. However, your life will probably be easier if you avoid using undefined in sequences, so you can test for it conclusively, and use null instead to represent missing data.

note about iterator.return()

In addition to a next() method, iterators may also implement a return() method: see overview from Exploring ES6

When you use a value-producing method, iter8 may not completely iterate over iterable objects it consumes. For example, if indexOf is successful, it will only iterate until it finds the element it's looking for.

The return() method can be implemented by iterables to allow them to clean up resources. iter8 will correctly invoke this method when needed whenever a query is executed.

In one situation, iter8 will not close an iterable that it consumes. This is when you actually source iter8 from an iterable (versus an iterator). Because iter8 was not resonsible for creating the iterable, it assumes it is also not responsible for cleaning it up. That is, the programmer could intend to take some further action on it after iter8 was done. An example:

const set = new Set([1, 2, 3, 4, 5])
let values = set.values();
const obj = iter(values).take(3).toArray()   // obj === [1, 2, 3]

let remainder = [];
for (let item of values) {
    remainder.push(item)
}
// remainder = [4, 5]

Because Set.values() returns an iterator object (rather than an iterable, from which iter8 would request a new iterator), it is considered to not be owned by iter8, and so it won't be closed after use.

Consider an alternate implementation:

const set = new Set([1, 2, 3, 4, 5])
let values = { 
    [Symbol.iterator]: function() { 
    	return set.values() 
    }
}

// ... rest code is the same

// remainder = [1, 2, 3, 4, 5]

In this case we created an actual iterable object (not an iterator) that returns the new sequence of keys. The output of "remainder" in this case will be the entire sequence. Each operation on values gets a new iterator from set.values() so the state is never preserved.

Creating Iter objects

iter8 exports a single function. This is a constructor, so it can be used for instanceof tests, but you can also simply invoke it rather than using new. To use iter8, create a new instance from an iterable object:

Create a new Iter from an iterable object:

import iter from `iter8`

let obj = iter()
// obj.toArray() === []
// obj instanceof iter === true 

let obj = iter([1,2,3]).filter(e => e<3)
// obj.toArray() === [2,3]

Map objects become sequences of [key, value] pairs:

let map = new Map(['foo', 'bar'], ['fizz,'buzz'])
let obj = iter(map)

// obj.toArray() === [['foo', 'bar'], ['fizz', 'buzz]]
// obj.toObject() === { foo: 'bar', fizz: 'buzz'}
// obj.as(Map) === Map { 'foo' => 'bar', 'fizz' => 'buzz' }

You can create an Iter directly from a generator (a function producing an iterator):

function* gen() {
    yield 1;
    yield 2;
    yield 3;
}

let obj = iter(gen);
/// obj.toArray() === [1,2,3]

You can can work with iterators directly, too:

let map = new Map([1, 'foo'], [2,'bar'], [3, 'baz'])

// map.values() returns an iterator - not an iterable!
let obj = iter(map.values())

// obj.toArray() === ['foo', 'bar', 'baz']

API Reference

iter(obj)

Create a new Iter instance from:

an iterable object. Iterable objects have a method [Symbol.iterator] that returns an iterator. Most JavaScript types that are containersimplement this, e.g. an Array, Map, Set, and arguments.
an iterator. Iterators are entities that have a method next(), and should implement the iterator protocol. Some built-in methods return iterators, like Map.values()
a generator. If you pass a function to iter, it is assumed to be a generator.

See iterator protocols for details on the specific entity types.

When iterable and generator types are used, each time the sequence for an Iter object is enumerated, the seqeunce will be restarted from the original source. That is, we will invoke source[Symbol.iterator]() for an iterable, and generator() for a generator, to obtain the seqeunce for a given enumeration.

With an iterator, however, we only have the sequence itself. We can't obtain it again. In this case, iter8 will cache the sequence the first time it is enumerated, and re-use the cached values for any subseqeuent enumeration. For this reason any operations on an Iter object sourced from an iterator are guaranteed to be reproducible, whereas operations on the others are not, since the sequence could be different each time it is requested. If you want to enforce reproducibility, then you should always cache a sequence at the beginning of an operation. See execute.

import iter from 'iter8'

const seq = iter([{value: 1},{ value: 2}, {value: 3}]);
const val = seq
    .filter(e=>e.value > 1)
    .sum('value');

// val === 5

const lookup = iter([['foo', 1], ['bar', 2]]).as(Map)
const val = lookup.get('bar')

// val === 2

iter.fromObject(obj, filter)

Create a new Iter of [key, value] pairs by enumerating properties on obj and its prototype chain. This accesses the same properties as for ... in, except ignores constructor.

You can pass a callback filter(prop) to filter properties. Returning false from the callback will exclude the property.

Example:

let obj = iter({ foo: 'bar', fizz: 'buzz'})
// obj.toArray() === [['foo', 'bar'], ['fizz', 'buzz]]

iter.fromObjectOwn(obj, filter)

Create a new Iter of [key, value] pairs, ignoring the prototype chain. This accesses the same properties as Object.keys.

You can pass a callback filter(prop) to filter properties. Returning false from the callback will exclude the property.

The default object creation behavior when passing an object directly to iter is the same as Iter.fromObjectOwn(obj) with no filter.

iter.generate(obj, n)

If obj is a function(n), invoke it with the index from 0 to 1-n, and create a sequence from each value returned.

If obj is not a function, create a sequence of obj repeated n times.

let x = iter.generate('foo', 3).concat('bar').toArray() 
// x === ['foo','foo','foo','bar']

let x = iter.generate((i)=>i*2, 3).toArray()
// x === [0,2,4]

iter.reflect(obj, recurse, filter)

Generate metadata about an object and optionally it's prototype chain (if recurse===true). Exclude methods based on a filter(name) callback.

reflect includes all regular properties as well as getters/setters. For the purposes of this documentation, a property is called a field if it's a regular property, and a property if it's a getter/setter.

This method returns a sequence of key-value pairs where the key is the property name, and the value is an object with properties:

type: the data type for fields ('string', 'number', 'function', 'object', 'null', 'undefined'). For properties the value is null. (Actuall null values are the string "null").
field: true for regular properties, false for getter/setter
writeable: boolean, the property can be assigned to
get: the property getter function or undefined
set: the property setter function or undefined
configurable: the property is configurable
enumerable: the property is enumerable
depth: the depth in the prototype chain. 0 means the object itself, 1 means its prototype, 2 means the prototype's prototype, etc.

Instance Methods

All the methods below are available on any Iter instance.

Transformation methods

These methods return a new sequence based on some transformation of the original one. These include aggregation, traversal, and set operations. These are all chainable, and execution is always deferred. The sequence will only be iterated as much as necessary to perform a given transformation. Some methods will by nature always iterate over the entire sequence, such as set operations.

flatten(recurse)

For each element in the sequence, if the element is an array, return each element in the array as a new element in the sequence.

When recurse is truthy, elements within each element that are also iterable, will continue to be flattened.

Like concat, strings are a special case, and are not iterated over.

let x = iter([[1,2], "foo", [3,4,[5]]]).flatten().toArray()

// x === [1,2,"foo",3,4,[5]]

Recursive:

let x = iter([[1,2], "foo", [3,4,[5]]]).flatten(true).toArray()
// x === [1,2,"foo",3,4,5]

unique(key)

Return a sequence containing only one element for each distinct value in the sequence.

key can be used to identify the value for equality comparison. With no arguments, the values themselves are made used. If a key is identified, then the first value in the sequence matching the key will be returned.

let x = iter([1,2,3,3,4,4,5,4,5]]).unique().toArray()
// x === [1,2,3,4,5]


let x = iter([ 
    { key: 1, value: 'foo'},
    { key: 1, value: 'foo-2'},
    { key: 2, value: 'bar'},
    { key: 3, value: 'fizz'},
    { key: 3, value: 'buzz'},
  ])
    .unique('key')
    .map(e=>e.value)
    .toArray()
// x === ['foo', 'bar', 'fizz']

groupBy(key, map)

Return a sequence of key-value pairs where each key is a distinct group, and each value is an Array of all the elements from the original sequence in that group.

key identifes the value on which to group. map is also a key argument which identifies a transform to be applied to each element from the input when it's added to a group.

let arr = [
    { category: 'home', value: 1}, 
    { category: 'work', value: 2},
    { category: 'home', value: 3},
]

let x = iter(arr).groupBy('category').toArray()
// returns [
//    ['home', [
//      { category: 'home', value: 1},
//      { category: 'home', value: 3}
//    ]],
//    ['work', [
//      { category: 'work', value: 2}
//    ]]
//  ]

Using a map:

let x = iter(arr).groupBy('category', 'value')toArray()
// returns [
//    ['home', [ 1,3 ],
//    ['work', [ 2 ] ]
//  ]

cast(Type)

Convert each element to an instance of Type. Type must be a constructor, and is invoked with the element as a single constructor argument for each element in the sequence.

A common use case for this is dealing with [key, value] pairs. Imagine a class Kvp that takes an array with two elements as its constructor argument, and returns an object exposing key and value properties. You could do things like this:

keys()

Assuming a sequence of [key, value] pairs, return a new sequence including only the keys. This is the same as map(0) but will perform basic type checking.

values()

Assuming a sequence of [key, value] pairs, return a new sequence including only the keys. This is the same as map(1) but will perform basic type checking.

map(callback(e, i), thisArg)

Transform each element to the return value of callback

filter(callback(e, i), thisArg)

Return a sequence including only elements that satisfy the condition in callback.

slice(begin, end)

Return a subset of the original sequence. skip and take will do the same thing, and may be more expressive. Negative values for "begin" not currently supported.

Merging and Set Operations

except(sequence)

Return only elements in the first sequence not found in the 2nd

let x = iter([1,2,3,4,5]]).except([3,5]).toArray()
// x === [1,2,4]

intersect(sequence)

Return only elements in found in both sequences

let x = iter([1,2,3,4,5]]).except([3,5]).toArray()
// x === [1,2,4]

union(sequence)

Return a sequence containing all unique elements found in either sequence.

If using an on clause to specify keys, for keys found in both sequences, the value from the original sequence will be returned in the resulting sequence.

let x = iter([1,2,3]]).union([2,3,4]).toArray()
// x === [1,2,3,4]

leftJoin(sequence, callback(leftItem, rightItem))

Join two sequences, and return a single new sequence.

The default behavior with no on clause assumes that the two sequences are [key, value] pairs with unique keys, e.g. Maps, and the key will be used to join them. It returns a new sequence, where the value is result of invoking mergeCallback(left, right, key) for the value of each matched entries.

You can provide an on clause to use sequences of any kind, including supporting sequences with duplicate keys.

There will be one row for each matching key in the right sequence. If keys are repeated in the left sequence, then each matching value in the right seqence will be repeated.

let seq1 = [[0,'foo'], [1,'bar'], [1,'baz'], [2, 'fizz']] 
let seq2 = [[1,'FOO'], [2,'BAR'], [3,'NOPE']]

let merged = iter(seq1)
.leftJoin(seq2, (left, right='') => `${left}:${right}`);

/// merged.toArray() =  ['foo:', 'bar:FOO', 'baz:FOO', 'fizz:BAR']

on(left: key, right: key)

Specify keys to use when performing operations that involve merging two sequences. This is valid only when it immediately follows one of these operations:

except
intersect
union
leftJoin

The two arguments will be used to obtain keys from the left and right sequences, respectively. The value that will be used to perform a set or join operation.

If only the left key is provided, then the same logic will be used on both the left and right sequences. If you wish to explicitly specify a key provider for only the left or right sequence, you can pass null or undefined as the other argument, and the values from that sequence will used directly as the key.

let seq1 = [
    { group: 1, value: 'bar' }, 
    { group: 1, value: 'foo', },
    { group: 2, value: 'fizz' },
    { group: 3, value: 'buzz' }
]
let seq2 = [
    { group: 1, value: 'a'},
    { group: 3, value: 'b'},
    { group: 3, value: 'c'},
    { group: 4, value: 'd'},
]

let inBoth = iter(seq1)
    .intersect(seq2)
    .on(e=>e.group);

// inBoth.toArray() = [{ group: 1, value: 'bar' }]

Note that the value returned by intersect is the value from the original sequence when using an on clause. The only time a value from the other sequence will appear when using on is in the case of a union, and it doesn't also appear in the original sequence.

let merged = iter(seq1)
    .leftJoin(seq2, (left, right={}, key)=> `${key}:${left.value},${right.value || ''}`
    .on(left => left.group, right => right.group)

// merged.toArray() ===["1:bar,a", "1:foo,a", "2:fizz,", "3:buzz,b", "3:buzz,c"]

Of note here is that the values from the left sequence group 3 are repeated, because there were multiple matches in the right group. The value for group 4 doesn't appear because it has no match from the left. Finally, we have to handle cases where right is missing in the merge callback, because there may be keys that appear only in the left sequence.

concat(obj, obj, ...)

Return a new sequence composed of the original sequence, followed by each element it the other sequences passed as arguments, or the other arguments themselves. Arguments do not need to be iterable; any non-iterable arguments will be appended to the sequence as a new element.

Strings are a special case. Despite being iterable, they are always appended to the sequence as a single element, consistent with Array.concat.

let x = iter([1,2,3]).concat([4,5,6], "foo", [7,8])
// x === [1,2,3,4,5,6,8 8]

Note that concat is not recursive; if an element of an appended sequence is itself iterable, it's not flattened.

let x = iter(['foo', 'bar']).concat(['baz', ['fizz'])
// x === ['foo','bar','baz',['fizz']

Ordering

orderBy(key)

Sort a sequence by comparing each element according to the value specified key.

let seq = [
    { foo: 2, id: 1}, 
    { foo: 1, id: 2}, 
    { foo: 3, id: 3}
]

let x = iter(seq).orderBy('foo').map((e)=>e.id).toArray();
// x = [2,1,3]

let x = iter(seq).orderBy(e=>e.foo).map((e)=>e.id).toArray();
// x = [2,1,3]

orderByDesc(key)

Same as orderBy, but sorts in descending order.

thenBy(key)

Chain a secondary (or n-ary) sort to an orderBy clause, which sorts orders which are equal during the primary sort, using the key.

let seq = [
    { foo: 2, bar: 2, id: 1}, 
    { foo: 1, id: 2}, 
    { foo: 3, id: 3},
    { foo: 2, bar: 1, id: 4}
]

let x = iter(seq)
    .orderBy('foo')
    .thenBy('bar')
    .map((e)=>e.id).toArray();

// x = [2,4,1,3]

If you attempt to use a thenBy clause anywhere other than directly after another sorting clause, an error will be thrown.

thenByDesc(key)

Same as thenBy, but sorts in descending order.

sort(callback(a, b))

Sort the sequence according to the relative comparison value returned by the callback. Same as Array.sort.

reverse()

Reverse the order of the sequence.

Traversal

skip(n)

Skip n elements in the sequence

let seq = iter([1,2,3,4,5])
let x = seq.skip(2).first()
// x === 3

let x = seq.skip(2).toArray()
// x === [3,4,5]

skipWhile(callback)

Skip elements in the sequence as long as the current element passes a test.

Each item is passed to the callback function and skipped as long as the element passes the test

let seq = iter([1,2,3,4,5])
let x = seq.skipWhile(e=>e < 4).first()
// x === 4

take(n)

Subset the sequence to include only the next n elements.

let x = iter([1,2,3,4,5]).skip(1).take(2).toArray() 
// x === [2,3]

Since each step operates against a new sequence defined by the previous step, successive take operations might operate counterintunitively -- e.g. x.take(3).take(2) is not the same as x.take(5) -- rather it's the same as x.take(2).

takeWhile(callback)

Subset the sequence to include items all items until the first item that fails a test.

Each item is passed to the callback function and included in the sequence until the first item fails the test.

let x = iter([1,2,3,4,5]).skipWhile(e=>e < 4) 
// x === [1,2,3]

Special

do(callback, thisArg)

do is similar to forEach in that it simply invokes a callback for each element in the sequence. Unlike forEach, do ignores the return value, and so cannot be canceled. The output sequence is always identical to the input sequence.

do is deferred, like every non-value-producing method, and so should not be used to immediately cause side effects like forEach.

execute()

Though this is not a value-producing method, it causes the query to be executed immediately, instead of waiting until a value-returning method is run. This is useful in cases where you want to keep your data in an Iter object, but cache an intermediate result for later use with other computations.

const intermediate = iter(something)
    .groupBy('category')
    .map([key, value]=>[key, iter(value).map(e=>e.amount).sum()])
    .execute();

let group1 = intermediate.filter([category]=>category === 'category 1').as(Map)
let group2 = intermediate.filter([category]=>category === 'category 2').as(Map)

Without using "execute" here, the "groupBy" etc. would be run twice for both group1 and group2 since execution of the entire query is deferred until a value-producing result, in this case as(Map).

Value-returning methods

These methods all cause the query to execute and return some value or object.

first(default)

Return the first element in the sequence. Same as get(0). If the sequence has no elements, return undefined, or the default if provided.

last(default)

Return the last element in the sequence. If the sequence has no elements, return undefined or default, if provided.

get(n, default)

Return the nth (0-based) element in the sequence.

Aggregation/Analysis methods

count()

Return the number of elements in the sequence.

let x = iter([1,2,3,4,5]).count()
// x === 5

min(key)

Return the minimum of all values in the sequence. If an optional key is provided, it will be used to produce the value to sum.

let x = iter([3,1,2,4]).min()
// x===1

max(key)

Return the max of all values in the sequence. If an optional key is provided, it will be used to produce the values to evaluate.

sum(key)

Return the sum of all values in the sequence. If an optional key is provided, it will be used to produce the values to evaluate.

mean(key)

Return the mean (average) of all values in the sequence. If an optional key is provided, it will be used to produce the values to evaluate.

some(callback(e, i), thisArg)

Test whether some elements in the sequence meet the criteria.

every(callback(e, i), thisArg)

Test whether every element in the sequence meet the criteria.

includes(value)

Test whether value is in the sequence

indexOf(value)

Locate value in the sequence and return it's ordinal position from 0, or -1 if not found.

lastIndexOf(value)

Locate the last occurrence value in the sequence and return it's ordinal position from 0, or -1 if not found.

findIndex(callback(e, i), thisArg)

Locate the element that satisfies the condition, and return it's ordinal position from 0. If the condition is never satisfied, return -1

find(callback(e, i), thisArg, default)

Attempt to locate an element in the sequence by using a callback, which should return true when the element has been located. If the condition is never satisfied, return undefined or default, if provided.

reduce(callback(last, current, i), initial)

(todo - Array.reduce)

reduceRight(callback(last, current, i), initial)

(todo - Array.reduceRight)

sequenceEqual(sequence)

Test whether the two sequences are equal by comparing each element sequentially. Returns true only if the sequences contain the same number of elements, and each element is equal.

Export methods

These methods export the sequence to some other data structure.

toArray()

Return an array created by iterating the sequence completely. Same as as(Array).

toObject()

Assuming the sequence contains key-value pairs and each key is unique, return an object with {property: value} for each key-value pair.

This method makes no effort to validate the elements in your sequence, rather, it just uses value[0] for each key and value[1] for each value. It is often useful in conjunction with groupBy, which returns a sequence of key/value pairs, or when iter is created from a Map object.

let myMap = new Map();
myMap.set('foo', 'bar')
myMap.set('fizz', 'buzz')

let x = iter(myMap).toObject();

// x === { 
//    "foo": "bar",
//    "fizz": "buzz"
//}

assignTo()

Assuming the sequence containst key-value pairs map them to the target object. Works the same as toObject() but assigns each kvp to an existing object rather than creating a new one. Returns the input object.

as(Type)

Creates an instance of Type using the sequence as a single constructor argument. This works well with the ES6 Map and Set types, or any other type constructed from an iterable.

You can also use Array or Object, which are special cases. These are aliases for the toArray() and toObject() methods.

// make a lookup table returning all elements of the source grouped by the
// values of property `category`

let map = iter(something).groupBy('category').as(Map)
let arr = iter(something).as(Array)

join(separator=",")

Join all methods using the separator as a string. If the contents of the sequence aren't strings, toString() will be invoked on each element.

Special Method

forEach(callback(e, i), thisArg)

forEach executes the query immediately, iterates over all values in the sequence, and always returns undefined. If you want to cause side effects during normal sequence processing, use do instead.

// log 'message' for all but the first 3 elements in the sequence
iter(arr).skip(3).forEach((e)=> {
    console.log(e.message)
});

Performance

While performance isn't my primary goal with this libary, I think some benchmarks are useful to understand how this performs overall and if it is adequate for your needs. I have some simple tests that compare it to lodash, which I consider the benchmark standard. iter8 will probably never be as fast as lodash for operations that involve iterating the entire sequence -- the purpose of these tests isn't to try to catch up, but just to compare.

Very roughly, whe starting with Array sources, typical complex operations that involve iterating the entire set seem to be about 3-5 times faster in lodash. This isn't likely to improve much, except to the extent that the Javascript engines can improve iteration access to arrays interally. It's just a fact of life since iterating over arrays is slower than accessing array elements by index.

This jsperf compares four methods of iterating an array. Using array indices is four times faster (in Chrome) than for-of -- which pretty much explains the performance difference compared to lodash. Other browsers show even more dramatic differences:

Edge is 100 times faster for indexed access
Firefox is 40 times faster for indexed access
Brave is about 15 times faster for indexed access

If your source is not already an Array, then iter8 may very well be faster, since you'd actually have to convert it to an array first to use with lodash. A simple test of "except" involving sources in Set objects shows iter8 to be about 20% faster.

At the end of the day - both lodash and iter8 are extremely fast. iter8 can sum 1,000,000 integers on my 2012 era Core i7 laptop in 0.05 seconds. Lodash can do it in .012 seconds! So if your needs are extremely computation heavy involving very large datasets, and eking out every bit of performance is critical, then lodash is certainly the best choice. But for most typical programming tasks, the difference is not likely to be perceptible.

Roadmap

A few more features

This is the only one common sequence method left to add that I can think of

zip(other, fn)

Apply a function to the corresponding elements of two sequences; return the output of the function for each element.

Typescript

I did a basic TypeScript conversion, but it doesn't actually work, and adds about 15% to the size anyway because of some optimizations that aren't directly possible with TS and Babel. But it's good enough to generate a typings file.

I am working independently on a good jsdoc-to-d.ts converter to provide typings (generally) for non-typescript projects. Right now I am manually maintaining the typings file, which is a bad idea.

Size Optimization

It's about 10K compressed/4k gzipped right now. Everything I see being part of it's purpose has already been implemented. There are definitely some opportunties to reduce the size a bit through better code reuse.

Similar Libraries

There are a bunch of them. It was actually pretty hard to find a good npm package name. :)

I had the following needs in mind when creating this:

easy interop with Array and iterables e.g. Map, Set, immutable library, and native Javascript objects
complete support of Array API
common sequence/set language and set and group operations
deferred execution of queries, and query re-use
good documentation
well tested
small

All the similar packages I found missed on one or more of these points.

iterable iterator generator array ES6 linq sets query

@everything-registry/sub-chunk-1940 @zalastax/nolb-ite

7 years ago

7 years ago

7 years ago

7 years ago