Awesome

florida

Pure functional accessor factories in js.

Warning: This library is not ready.

Some parts of the documentation are based on a previous incarnation.

Work in progress

Install (TBD)

npm install florida

Testing

Install npm and bower dependencies:

npm install
npm test

Summary of API

Hypercubed/florida is a shortcut for composable data accessors functions. For example:

Import

ES6

import * as F from 'florida';
import {keys as FK} from 'florida';

CJS

var F = require('florida');
var FK = F.keys;

Accessors

Example

var data = [ { firstname: 'John', lastname: 'Smith', age: 51 }, /* ... */ ];
var firstname = FK('firstname');

data.map(_firstname.get());  // Returns a list of first names

Operators

Example

var _johns = _firstname.eq('John');

data.filter(_johns);  // returns a list of John's

Chaining

Example

var _age = FK('age');
var _twenties = _age.both(FK().gte(20), FK().lt(30));

data.filter(F.and(_johns, _twenties));  // returns a list of John's in their twenties

Sorting

Example

data.filter(FK.and(_johns, _twenties)).sort(_age.asc());  // returns a list of John's in their twenties sorted by age in ascending order

Why?

In JavaScript, especially when using d3, we often write accessor functions like this:

function(d) { return d.value; }

This simple function returns the value of the value key when an object is pass to it. For example in the map function:

values = data.map(function(d) { return d.value; });

This is lightweight, simple, and readable. There is nothing wrong with it. Sometimes, however, in order to avoid repeating ourselves so we crete a reusable accessor function like this:

var _value = function(d) { return d.value; };
values = data.map(_value);

Now imagine the object also has a year key whose values are date objects. We may want to filter like this:

var _value = function(d) { return d.value; };
var _year_filter = function(d) { return d.year >= new Date('1980 Jan 1'); };
values = data.filter(_year_filter).map(_value);

However, this has a couple of slight drawbacks. First of all you will need to create a new filter every time the date changes; also the Date constructor is called for every element in the data array. A better approach is an accessor factory:

var _year_filter = function(date) {
  return function(d) { return d.year >= date; };
}

var _filter = _year_filter(new Date('1990 Jan 1'));
values = data.filter(_filter).map(_value);

It's a little ugly but here the Date constructor is only called once and the _year_filter function returns the accessor. An new accessor can be created any time by calling _year_filter

Now what if we want to filter between two dates. We can do modify our accessor factory:

var _year_filter = function(dateA, dateB) {
  return function(d) { return d.year >= new Date(dateA) && d.year < new Date(dateB); };
}

but let's say that you have multidimensional data where dateA and dataB change independently. You might be tempted to do something like this:

var _year_gte = function(dateA) {
  return function(d) { return d.year >= dateA; };
}

var _year_lt = function(dateB) {
  return function(d) { return d.year < dateB; };
}

_year_filter1 = _year_gte(new Date('1980 Jan 1'));
_year_filter2 = _year_lt(new Date('1990 Jan 1'));

values = data
  .filter(_year_filter1)
  .filter(_year_filter2)
  .map(_value);

Ok, no we are getting ridiculous. The date constructor is not that expensive. But you can imagine a situation where the values for filters could be very expensive. For example based on aggregated statistics or reading from the DOM.

Ok, at this point let me introduce FK. FK is simply a shortcut for all this. For example:

var _value = FK('value');
values = data.map(_value.$);

The value returned from FK('value').$ in this case is simply the accessor function function(d) { return d.value; }.

Interesting. How about this:

var _value = FK('value');
var _year_filter = FK('year').gte(new Date('1980 Jan 1'));
values = data.filter(_year_filter).map(_value.$);

FK('year').gte(somevalue) is essentially a shortcut for function(d) { return d.year >= somevalue; }.

It gets better:

var _value = FK('value');

var _year_filter =
  FK('year').both(FK().gte(new Date('1980 Jan 1')), FK().lt(new Date('1990 Jan 1')));

values = data.filter(_year_filter).map(_value.$);

or how about this:

var _value = FK('value');
var _value_filter = _value.gt(10);

var _year = FK('year');
var _year_filter = _year.both(FK().gte(new Date('1980 Jan 1'), FK().lt(new Date('1990 Jan 1')));

var _filter = F.and(_value_filter, _year_filter);

values = data.filter(_filter).map(_value.$);

Pretty neat?

Acknowledgments

License

Copyright (c) 2015 Jayson Harshbarger MIT