Object-Scan
![Gardener](https://github.com/blackflux/js-gardener/blob/master/assets/badge.svg)
Find keys in object hierarchies using wildcard matching and callbacks.
Install
Install with npm:
$ npm install --save object-scan
Usage
const objectScan = require('object-scan');
objectScan(['a.*.f'])({ a: { b: { c: 'd' }, e: { f: 'g' } } });
Features
- Object and Array matching with e.g.
key.path
and [1]
- Wildcard matching with
*
and [*]
- Partial Wildcard matching with e.g.
mark*
, m?rk
, [*1]
or [?1]
- Arbitrary depth matching with
**
- Or-clause with e.g.
{a,b}
and [{0,1}]
- Exclusion with e.g.
!key
- Full support for escaping
- Input traversed exactly once during search
- Results returned in "delete-safe" order
- Search syntax is checked for correctness
- Dependency free, small in size and very performant
- Lots of tests to ensure correctness
Options
Signature for all functions is
Fn(key, value, { parents, isMatch, matchedBy, excludedBy, traversedBy })
where:
key
is the key that the function is called for (respects joined
option).value
is the value for that key.parents
is array of form [parent, grandparent, ...]
.isMatch
is true if the last targeting needle exists and is non-excluding.matchedBy
are all non-excluding needles targeting the key.excludedBy
are all excluding needles targeting the key.traversedBy
are all needles involved in traversing the key.
filterFn
Type: function
Default: undefined
If function is defined, it is called for every match. If false
is returned, the current key is excluded from the result.
Can be used as a callback function to do processing as matching keys are traversed.
Called in same order as matches would appear in result.
This method is conceptually similar to the callback function in
Array.filter().
breakFn
Type: function
Default: undefined
If function is defined, it is called for every key that is traversed by
the search. If true
is returned, all keys nested under the current key are
skipped in the search and from the final result.
Note that breakFn
is called before the corresponding filterFn
might be called.
joined
Type: boolean
Default: true
Can be set to false to return each key as a list. When dealing with special characters this can be useful.
Setting this to false
improves performance.
Note that _.get and _.set fully support lists.
useArraySelector
Type: boolean
Default: true
When set to false
, no array selectors should be used in any needles and arrays are automatically traversed.
Note that the results still include the array selectors.
strict
Type: boolean
Default: true
When set to false
, no errors are thrown when:
- the input contains redundant needles
Examples
More extensive examples can be found in the tests.
const objectScan = require('object-scan');
const obj = {
a: {
b: {
c: 'd'
},
e: {
f: 'g'
},
h: ['i', 'j']
},
k: 'l'
};
objectScan(['*'])(obj);
objectScan(['a.*.f'])(obj);
objectScan(['*.*.*'])(obj);
objectScan(['a.*.{c,f}'])(obj);
objectScan(['a.*.{c,f}'], { joined: false })(obj);
objectScan(['*.*[*]'])(obj);
objectScan(['*[*]'])(obj);
objectScan(['**'])(obj);
objectScan(['**.f'])(obj);
objectScan(['**[*]'])(obj);
objectScan(['a.*,!a.e'])(obj);
objectScan(['**'], { filterFn: (key, value) => typeof value === 'string' })(obj);
objectScan(['**'], { breakFn: (key) => key === 'a.b' })(obj);
Edge Cases
The top level object(s) are matched by the empty needle ""
.
Useful for matching objects nested in arrays by setting useArraySelector
to false
.
Note that the empty string does not work with _.get and _.set.
Special Characters
The following Characters are considered special and need to
be escaped if they should be matched in a key: [
, ]
, {
, }
, ,
, .
, !
, ?
and *
.
When dealing with special characters, it might be desirable to set the joined
option to false
.
Internals
Conceptually this package works as follows:
-
During initialization the needles are parsed and built into a search tree.
Various information is pre-computed and stored for every node.
Finally the search function is returned.
-
When the search function is called, the input is traversed simultaneously with
the relevant nodes of the search tree. Processing multiple search tree branches
in parallel allows for a single traversal of the input.
Having a separate initialization stage allows for a performant search and
significant speed ups when applying the same search to different input.