使用indexOf使用Ramda（或其他功能库）过滤大型对象数组。

Question

I love Ramda, but sometimes i stumble in seemingly simple things. What i have, is a list of places and words i need to look into (their indexOf must be equal to 0) to generate a list of search suggestions, so for example, given the list below, i would like to suggest to a user all 3 areas (by name) when they type 'aber' in the search box but only Aberdeen North if they type 'nor'.

[
  {
    name: "Aberavon",
    words: [ "aberavon" ]
  },
  {
    name: "Aberconwy",
    words: [ "aberconwy" ]
  },
  {
    name: "Aberdeen North",
    words: [ "aberdeen", "north" ]
  },
]

I managed to get the 'proper' thing running getting my results through console log, however i do not seem to get the filtered items. My code looks like this and the console logs will correctly show the iteration going through the words and the indexOf matching the start of a term but i never get my objects back out of the filter. I tried to make it even more functional and even more non-functional to no result so far! This is my code:

var matches = R.filter ( function ( x ) { R.map ( function ( word ) {
        console.log ( word );
        console.log ( word.indexOf ( term ) === 0 );
        R.equals ( word.indexOf ( term ), 0 )
    } ) ( x.words ) } ) ( json );

Answer 1

Not a Ramda expert, but I might be able to help.

The function (first argument) you pass to R.filter should take an object and return a bool .

The key is to create this method in a readable and reusable way. Eventually, you'll end up with something like:

• Go from an object {} to an array of strings [string]
• Go from an array of strings [string] to a boolean bool

Step 1: Checking if a word matches a term

Currently, you have defined a function that is basically your "starts with" test:

R.equals ( word.indexOf ( term ), 0 )

This function needs two strings to work: word and term . (Technically they could be anything that implements indexOf , but let's stick to the example)

I'd start by testing this method and giving it a name, so you know this part is "done" and works.

const startsWith = term => word => word.indexOf(term) === 0;

(Later, you can rewrite this to use the Ramda API and include other features like case (in)sensitivity. You might also want to annotate something like string -> string -> bool , but again, I don't know the Ramda way)

Step 2: Checking if some string matches a term

Now that you can check if strings match a term, you'll need to find out if at least one string in an array of strings matches a term.

In plain javascript:

const anyStartsWith = term => arr => arr.some(startsWith(term));

I think the Ramda equivalent is R.any :

const anyStartsWith = term => R.any(startsWith(term));

Again, test this method and see if it performs like you want it to.

Step 3: Check if a Place matches a term

This is the most complicated step. We need to go from an object with a words property to something our earlier defined filter methods can handle.

const placeMatchesTerm = term => place =>
  anyStartsWith(term) (place.words);

Step 4: Filtering

Now, we have a function that takes a term and returns a function that takes a Place . This, we can use to filter our array of places directly:

const myPlaces = [ /* ... */ ];
const filter = (term, arr) => 
  arr.filter(placeMatchesTerm(term));

const aber = filter("aber", myPlaces);
const nor = filter("nor", myPlaces);

In a working example (without Ramda)

 // string -> string -> bool const startsWith = term => word => word.indexOf(term) === 0; // string -> [string] -> bool const anyStartsWith = term => arr => arr.some(startsWith(term)); // string -> {} -> bool const placeMatchesTerm = term => place => anyStartsWith(term) (place.words); // string -> [{}] -> bool const filter = term => arr => arr.filter(placeMatchesTerm(term)); const aber = filter("aber")(getPlaces()); const nor = filter("nor")(getPlaces()); console.log("'Aber' matches", aber.map(p => p.name)); console.log("'Nor' matches", nor.map(p => p.name)); // Data function getPlaces() { return [{name:"Aberavon",words:["aberavon"]},{name:"Aberconwy",words:["aberconwy"]},{name:"Aberdeen North",words:["aberdeen","north"]}]; } 

In Ramda, it might be something like this , but again, I'm no expert. I'll add the Ramda tag to your question which makes it pretty likely someone will show up to show you the Ramda ways :)

Answer 2

I like the previous answer quite a bit, but if you want to see how a Ramda person does it, this would be my solution:

const startsWith = R.curry((term, word) => word.indexOf(term) === 0);
const anyStartsWith = R.curry((term, words) => R.any(startsWith(term), words));
const anyWordMatches = R.curry(
  (term, place) => anyStartsWith(term, R.prop('words', place))
);
const anyPlaceMatches = R.curry(
  (term, places) => R.filter(anyWordMatches(term), places)
);

anyPlaceMatches('nor', json); 
//=> [ <Aberdeen North> ]
anyPlaceMatches('aber', json); 
//=>[ <Aberavon>, <Abercrombie>, <Aberdeen North> ]

While all of this can probably be made point-free, I don't see the, ahem, point of that.

You can see all this in the Ramda REPL .