2018-09-28 13:59:05 +03:00
# walk.js
2018-09-28 15:36:27 +03:00
2018-09-28 13:59:05 +03:00
An extensible tree walk(..) framework...
2018-09-28 15:36:27 +03:00
2018-09-29 23:44:44 +03:00
2018-09-28 15:36:27 +03:00
## Theory and operation
2018-09-29 23:44:44 +03:00
This module generalizes structure traverse (*walking*). This is done via a `walk(..)` function that recieves a user-defined `getter(..)` function and returns a *walker* .
### Constructing the walker and walking
2018-09-28 15:36:27 +03:00
`walk(getter)(state, ...nodes) -> state`
`walk(getter, state)(...nodes) -> state`
`walk(getter, state, ...nodes) -> state`
- Recieves a `getter` function a `state` and a list of `nodes` ,
- Iterates through `nodes` calling the `getter(..)` per node, threading the `state` through each call,
- Returns the `state` when there are no more `nodes` .
2018-09-29 23:44:44 +03:00
2018-09-30 00:16:09 +03:00
### Getting and processing nodes
2018-09-29 23:44:44 +03:00
2018-09-30 00:16:09 +03:00
`getter(state, node, next, stop) -> state`
- Recieves `state` , `node` and two control functions: `next` and `stop` ,
2018-09-29 03:13:46 +03:00
- Called in a context (`this` ), persistent within one `walk(..)` call, inherited from *walker* `.prototype` and usable to store data between `getter(..)` calls,
2018-09-28 15:36:27 +03:00
- Can process `node` and `state` ,
2018-09-30 00:16:09 +03:00
- Can queue nodes for walking via `next('queue', state, ...nodes)`
- Can walk nodes directly via `next('do', state, ...nodes) -> state`
2018-09-28 15:36:27 +03:00
- Can abbort *walking* and return a state via `stop()` or `stop(state)`
- Returns `state` ,
2018-09-29 23:44:44 +03:00
### Putting it all together
2018-09-28 15:36:27 +03:00
A trivial *flat* example...
```javascript
walk(function(r, n){ return r+n }, 0, ...[1,2,3]) // -> 6
```
The above is essentially equivalent to...
```javascript
[1,2,3].reduce(function(r, n){ return r+n }, 0) // -> 6
```
2018-09-29 23:44:44 +03:00
And for *flat* lists `.reduce(..)` and friends are simpler and more logical. `walk(..)` is designed to simplify more complex cases:
2018-09-28 15:36:27 +03:00
- The input is not *flat* :
```javascript
2018-09-29 23:44:44 +03:00
// sum the items in a *deep* array (depth-first)...
2018-09-30 00:16:09 +03:00
var sum = walk(function(r, n, next){
2018-09-28 15:36:27 +03:00
return n instanceof Array ?
2018-09-30 00:16:09 +03:00
next('queue', r, ...n)
2018-09-28 15:36:27 +03:00
: r + n }, 0)
sum( [1, [2, 3], 4, [[5], 6]] ) // -> 21
```
2018-09-30 00:16:09 +03:00
For reference here is a *recursive* `.reduce(..)` example, already a bit more complex:
2018-09-28 15:47:58 +03:00
```javascript
function sumr(l){
return l.reduce(function(r, e){
return r + (e instanceof Array ?
sumr(e)
: e) }, 0) }
sumr( [1, [2, 3], 4, [[5], 6]] ) // -> 21
```
2018-09-29 23:44:44 +03:00
2018-09-28 15:36:27 +03:00
- Need to abort the recursion prematurelly:
```javascript
2018-09-28 15:47:58 +03:00
// check if structure contains 0...
2018-09-30 00:16:09 +03:00
var containsZero = walk(function(r, e, next, stop){
2018-09-29 03:13:46 +03:00
// NOTE: we'll only count leaf nodes...
this.nodes_visited = (this.nodes_visited || 0)
2018-09-28 15:36:27 +03:00
return e === 0 ?
2018-09-30 00:16:09 +03:00
// abort search, report number of nodes visited...
2018-09-29 03:13:46 +03:00
stop(this.nodes_visited+1)
2018-09-28 15:36:27 +03:00
: e instanceof Array ?
2018-09-30 00:16:09 +03:00
next('queue', ...e)
2018-09-29 03:13:46 +03:00
: (this.nodes_visited++, r) }, false)
2018-09-28 15:36:27 +03:00
2018-09-29 03:13:46 +03:00
containsZero( [1, [2, 0], 4, [[5], 6]] ) // -> 3
2018-09-28 15:36:27 +03:00
containsZero( [1, [2, 5], 4, [[5], 6]] ) // -> false
```
2018-09-28 15:47:58 +03:00
See a more usefull search in [examples ](#examples )...
2018-09-28 15:36:27 +03:00
## Installation and loading
```shell
$ npm install --save generic-walk
```
```javascript
var walk = require('generic-walk').walk
```
*Note: This module supports both AMD and node's `require(..)` **
## API
`walk(getter(..)) -> walker(state, ...nodes)`
Construct a reusable walker.
`walk(getter(..), state) -> walker(...nodes)`
Construct a reusable walker with fixed initial state.
`walk(getter(..), state, ...nodes) -> result`
Walk the nodes.
### The getter
2018-09-30 00:16:09 +03:00
`getter(state, node, next(..), stop(..)) -> state`
2018-09-28 15:36:27 +03:00
User provided function, called to process a node.
2018-09-30 00:16:09 +03:00
`next('queue', state, ...nodes) -> state`
Queue `nodes` for walking. The queued nodes will get *walked* after this level of nodes is done (i.e. the `getter(..)` is called for each node on level). `state` is returned as-is. This is done to make `next('queue', ..)` and `next('do', ..)` signature compatible and this simpler to use.
2018-09-28 15:36:27 +03:00
2018-09-30 00:16:09 +03:00
`next('do', state, ...nodes) -> state`
2018-09-28 15:36:27 +03:00
Walk `nodes` and return `state` . The nodes will get *walked* immidiately.
`stop()`
`stop(state)`
Stop walking and return `state` . The passed `state` is directly returned from the *walker* .
*Note that `stop(..)` behaves in a similar manner to `return` , i.e. execution is aborted immidiately.*
## Examples
Sum all the values of a nested array (breadth-first)...
```javascript
var sum = walk(function(res, node, next){
return node instanceof Array ?
2018-09-30 00:16:09 +03:00
next('queue', res, ...node)
2018-09-28 15:36:27 +03:00
: res + node }, 0)
sum([1, [2], 3, [[4, 5]]]) // -> 15 ...walks the nodes: 1, 3, 2, 4, 5
```
Sum all the values of a nested array (depth-first)...
```javascript
2018-09-30 00:16:09 +03:00
var sumd = walk(function(res, node, next){
2018-09-28 15:36:27 +03:00
return node instanceof Array ?
2018-09-30 00:16:09 +03:00
// yes, this is the only difference...
next('do', res, ...node)
2018-09-28 15:36:27 +03:00
: res + node }, 0)
sumd([1, [2], 3, [[4, 5]]]) // -> 15 ...walks the nodes: 1, 2, 3, 4, 5
```
2018-09-29 23:44:44 +03:00
To explicitly see the paths the `sum` /`sumd` take we need to modify them a little:
2018-09-28 15:36:27 +03:00
```javascript
2018-09-30 00:16:09 +03:00
var makeSummer = function(mode){
return walk(function(res, node, next){
this.log(node)
return node instanceof Array ?
next(mode == 'breadth-first' ? 'queue' : 'do', res, ...node)
: res + node }, 0) }
var sum = makeSummer('breadth-first')
var sumd = makeSummer('depth-first')
2018-09-28 15:36:27 +03:00
2018-09-29 23:44:44 +03:00
// define the path logger...
2018-09-29 03:13:46 +03:00
sum.prototype.log =
2018-09-29 23:44:44 +03:00
sumd.prototype.log =
function(node){
this.path = node instanceof Array ?
this.path
: (this.path || []).concat([node])
}
// XXX need a more natural way to catch the end of the walk...
sum.prototype.onWalkEnd =
sumd.prototype.onWalkEnd =
function(res){
console.log('-->', this.path)
return res
}
2018-09-29 03:13:46 +03:00
sum([1, [2], 3, [[4, 5]]]) // -> 15
2018-09-28 15:36:27 +03:00
2018-09-29 03:13:46 +03:00
sumd([1, [2], 3, [[4, 5]]]) // -> 15
```
2018-09-28 15:36:27 +03:00
FInd first zero in tree and return it's path...
```javascript
2018-09-29 03:13:46 +03:00
// NOTE: the only reason this is wrapped into a function is that we need
// to restrict the number of items (L) this is passed to 1...
var firstZero = function(L){
2018-09-30 00:16:09 +03:00
return walk(function(res, node, next, stop){
2018-09-29 03:13:46 +03:00
// setup...
if(this.path == null){
this.path = []
node = [null, node]
}
var path = this.path
var k = node[0]
var v = node[1]
return v === 0 ?
// NOTE: we .slice(1) here to remove the initial null
// we added in firstZero(..)...
stop(path.slice(1).concat([k]))
: v instanceof Object?
(path.push(k),
2018-09-30 00:16:09 +03:00
next('do', res, ...Object.entries(v)))
2018-09-29 03:13:46 +03:00
: res}, false, L) }
2018-09-28 15:36:27 +03:00
firstZero([10, 5, [{x: 1, y: 0}, 4]]) // -> ['2', '0', 'y']
```