Logipar

/lɑːdʒɪpɜːr/

noun: Your go-to polyglot logic string parser.

verb: Parse that logic string, my friend.

What is a logic string parser?

Have you ever wanted to filter data based on a string of ANDs and ORs and NOTs? Well now you can!
In fact, that's exactly what Logipar is here to help you do.

If that doesn't help, check out our Cat breed demo to see Logipar in action! Or you can try it yourself on Runkit.

Logipar supports:

• AND
• OR
• NOT
• XOR
• and Parenthises

You can rename the operatores however you want! You can also use it with basically whatever literals/values you want¹. (1. There are some restrictions.)

A simple example: one AND (two OR three)
A more complex example: title=Cat XOR title contains dog. Logipar doesn't care about the literals you use, so you can add whatever complexity you is appropriate for your project there - I'm not judging! In fact, I endorse it.

Logipar will automatically merge adjacent literals, unless you don't want it to. So title contains dog can count as a single literal, even without quotation marks. Or it can count as three title, contains, and dog - the choice is yours! (By default it merges.)

What can I do with it once it's parsed?

Oh, man, whatever you want really! You can test objects against the parse tree (using a function Logipar returns - see the filtering section). You can flatten the parse tree to a string of your design (with the help of Logipar's stringify() function - see the stringing section).
I guess.. I guess that's really all. But come on, what more do you even want??

Okay cool, can I use it though?

Logipar is written in Haxe. WAIT DON'T BE SCARED. All that means for you is it can be (and is) compiled to multiple languages. Maybe you've heard of one of these before:

• Javascript
• Python
• PHP

There are more, but I'm going to focus the examples on those for now.

Installation

Javascript

• npm i logipar
• yarn add logipar
• Or you can just download and use Logipar.js - say in a <script src="Logipar.js"></scrip> tag.

Python

• pip install logipar
• Or you can just download and use Logipar.py in your project.

PHP

Bad news. Logipar isn't on Composer. I probably won't bother to add it there unless someone really wants it, because of this issue. Good news! You can totally use this Logipar.phar instead. Or, you know, donwload the files directly if you want.

Haxe

• haxelib install logipar
• Also the source files are here.

But how do I use it?

Great question! Here are some examples in different languages.

Usage

Javascript

You can follow along with these javascript samples on on Runkit.

const logipar = require("logipar")
const lp = new logipar.Logipar();
lp.parse("a AND b");
console.log(lp.toString())

Or you can include Logipar.js in your code. Note that in this method the classes are accessed via Logipar, Token, and Node - rather than through a logipar constant, as above.

<script src="Logipar.js"></script>

    // Include the library however works for you.  You can see how I did it in js_sample.html
    var lp = new Logipar();
    lp.parse("a AND b");
    console.log(lp.stringify());

Python

    import logipar
    lp = logipar.Logipar()
    lp.parse("a AND b")
    print(lp.toString())

Php

When you're using the PHAR, it should take care of loading the classes for you.

    require_once("Logipar.phar");   
    $lp = new \logipar\Logipar();
    $lp->parse("a AND b");
    print($lp->toString());

Quotations marks

While Logipar can automatically merge neigbouring literals, sometimes that's not enough. It also supports quotation marks around literals. This means you can have values that would otherwise be parsed as tokens in the literals - if they're wrapped in quotation marks. The quotation marks become part of the value, for you to deal with however you want. This is also handy for supplying whitespace as a value. Take the logic string a="CAT OR DOG" OR This is a sentance.. a="CAT OR DOG" is a literal. Even though it has and an OR in it. You can then use it however is appropriate - split it on the equals sign maybe, and strip the quotation marks to check for the string "CAT OR DOG" in the "a" column. I don't know, that's your journey!

This is a sentance. is also a literal since it'll be automatically merged by default. Logipar's mergeAdjacentLiterals (which defaults to true) controls this. So set it to false if you don't want to merge them.

So it'll parse out to: a="CAT OR DOG" OR This is a sentance..

The default quotation mark characters are " and '. But you can add to or change these through Logipar's aptly-named quotations property - which is an array of strings denoting whatever you want to use as valid quotation mark characters.

Javascript

lp.quotations.push("`"); // Add backtick 
lp.mergeAdjacentLiterals = true; // This is its default value

Python

lp.quotations.append("`"); // Add backtick 
lp.mergeAdjacentLiterals = true; // This is its default value

Php

$lp->quotations[] = "`"; // Add backtick
$lp->mergeAdjacentLiterals = true; // This is its default value

Case sensitivity

By default the Logipar's operators are case-sensitive, but they don't have to be. Simply change the caseSensitive property to false, and YoU cAn TyPe ThEm HoWeVeR yOu Want!

Javascript

lp.caseSensitive = false;

Python

lp.caseSensitive = False;

Php

$lp->caseSensitive = false;

Custom operators

You can also replace the default strings for any or all of logipar's operators. Maybe you want to go the old ^ v route. Or maybe && and ||. Or even just + and *. I don't know what you want! Valid operators are:

• Token.AND default: AND
• Token.OR default: OR
• Token.XOR default: XOR
• Token.NOT default: NOT
• Token.OPEN default: (
• Token.CLOSE default: )

You should probably keep the OPEN and CLOSE (parentheses) operators as single characters, unless you want to enforce whitespace between all tokens tokens.

Javascript

lp.overwrite(logipar.Token.AND, "&&");

Python

lp.overwrite(logipar.logipar_Token.AND, "et")

Php

PHP is a little more problematic. AND, OR and XOR are keywords in it, which makes it difficult to access those Token constants. Rather than rename those variables, you can just use the string values. (You can do the same with the other ones, if you really want to - all the string values are the same as the constant after Token..)

$lp->overwrite("AND", "et");

Stringification

Sometimes you want your logic tree flattened; pressed firmly into a string. Maybe you just want to display it, or maybe you'd like to use it in your SQL. I don't know - and I'm not judging. Logipar should provide for all your stringifying needs with it's stringify() function. When you call stringify, you have the option of passing a function to it - this function is used to convert nodes to strings in any manner you like. It will be called on each node in the tree. Anything you don't account for will use the default toString() function provided by Logipar.

That's confusing right? Well here, take a look at this function in Haxe:

function mystringer(n:logipar.Node):String {
	if (n.token.type == logipar.Token.XOR) {
		return "((" + n.f(n.left) + " AND NOT " + n.f(n.right) + ") OR (NOT " + n.f(n.left) + " AND " + n.f(n.right) + "))";
	}
	return null;
}

There are some things going on there. Definite things. Let's start with the signature: Your function should take a logipar.Node as a param, and return a String.
Say whatever we're using doesn't support an XOR operation. That's okay, a XOR b is just a fancy way of saying (a AND NOT b) OR (NOT a AND b). We can handle that. First, we need to know if this is the type of node we want to change (XOR), so we check if the node passed in (n) is currently of type logipar.Token.XOR. Then we just return the string the way we want it. In this case: return "((" + n.f(n.left) + " AND NOT " + n.f(n.right) + ") OR (NOT " + n.f(n.left) + " AND " + n.f(n.right) + "))";

But wait. What's going on there? Well XOR nodes (and all binary nodes) have left and right properties, representing their preceeding and succeeding operands. (Unary nodes like NOT only have a right property; or rather, left will be null.) So we're just saying: (({LEFT} AND NOT {RIGHT}) OR (NOT {LEFT} AND {RIGHT}))

With one added wrinkle. The left and right properties are nodes themselves. They may contain XORs of their own. So we want to recursively call the same stringification function on them. f() is a helper function available for the duration of the stringification process for this very purpose. That's why you see n.f(n.left) above.

The return null; lets Logipar know it should display any other node as usual. So in this case, anything that's not an XOR gets displayed as it normally would.

Javascript

var str = lp.stringify(function(n) {
	if (n.token.type == logipar.Token.XOR) 
		return "((" + n.f(n.left) + " AND NOT " + n.f(n.right) + ") OR (NOT " + n.f(n.left) + " AND " + n.f(n.right) + "))";
	return null;
});

Python

def expandXOR(n):
	if n.token.type == logipar.logipar_Token.XOR:
		l = n.f(n.left)
		r = n.f(n.right)
		return "(({} AND NOT {}) OR (NOT {} AND {}))".format(l, r, l, r)
	return None

flattened = lp.stringify(expandXOR)

Php

Note that again I'm just using the string "XOR" in PHP. Also, I'm using call_user_func to call $n->f() on the child nodes.

$flattened = $lp->stringify(function($n) {
	if ($n->token->type == "XOR") {
		$l = call_user_func($n->f, $n->left);
		$r = call_user_func($n->f, $n->right);
		return "((" . $l . " AND NOT " . $r . ") OR (NOT " . $l . " AND " . $r . "))";
	}
	return null;
});

The string returned will make use if minimal parentheses. If for some reason you want everything wrapped in brackets, that's easy too:

lp.walk(function(n:logipar.Node):Void { n.bracketing = logipar.Node.MAXIMAL_BRACKETS; });

That'll set the bracketing mode for each node in the tree to MAXIMAL_BRACKETS. If you only want to change certain node types, you can check the value in n.token.type and act accordingly.

Filtering data

Sometimes you just want to filter an array of rows. Nothing more, nothing less. Well, maybe more. Maybe you want to do it based on a logic string. Logipar's filterFunction can help. It creates a function you can use to filter your data. But how does it work? You handle the leaves, and we'll handle the logic tree.

Basically, you just need to decide if a given leaf resolves true or false for a given row of data. And then we'll figure out if it matches overall.

Here's an example in Haxe:

var leafresolver = function(row:Dynamic, value:String):Bool {
	// This is  just checks a leaf node (value) against every column in the data (row), in a case-insensitive way.  
	// But you can get as complex as you'd like and parse the value variable however you like.
	for (f in Reflect.fields(row)) { // For each property of row
		if (Std.string(Reflect.field(row, f)).toLowerCase().indexOf(value.toLowerCase()) != -1) // If that property contains the leaf 
			return true;
	}
	return false;
}
var myfilter:(Dynamic)->Bool = ls.filterFunction(leafresolver);

Okay, so you can see above that Logipar's filterFunction() takes a function as its argument, and returns a function. The first function (leafresolver()) we supply, the second we use to actually do our filtering. leafresolver takes a row of data. This is probably an object of some sort, but that's your journey. For the sake of our example, let's say it's {title: "Harry potter", "author": "J.K. Rowling"}. It also takes a string value. This is the value of the leaf we're checking. For the sake of our example, harry.

The task of this function is to take value and see if it matches for row. You can do this however you want. This function is then run on every LITERAL (the leaves of the logic tree), and we use its result to decide if the logic tree resolves to true or false for row.

For this example, it'd check Har against each property in row: title, and author. Since the title is Harry Potter, and we've specified in the function to convert to lowercase before checking, it'll match and return true.

filterFunction returns a function, whcih you can then use on your data. For example, myfilter(data[i]) will return true or false depending on if it matches the logic of the query.

That's still pretty confusing, but hopefully some more examples will clear it up.

Javascript

function leafresolver(row, value) {
	// This is  just checks the values against every column, in a case-insensitive way
	for(var field in row)
		if (row[field].toString().toLowerCase().includes(value.toLowerCase()))
			return true;
	return false;
}
f = lp.filterFunction(leafresolver);
filtered_data = sample_data.filter(f);    // Javascript arrays have a filter function

Python

def leafresolver(row, value):
	# This is  just checks the values against every column, in a case-insensitive way
	for field in row:
		if value.lower() in str(row[field]).lower():
			return True
	return False

f = lp.filterFunction(leafresolver)
data = list(filter(f, data)) # Python has a filter function too

Php

$leafresolver = function($row, $value) {
	foreach($row as $field=>$v)
		if (stripos($row[$field], $value) !== false)
			return true;
	return false;
};

$f = $lp->filterFunction($leafresolver);
$data = array_filter($data, $f);    // Oh look, so does PHP

Now, let's try a more complex example in Haxe:

var f = ls.filterFunction(function(row:Dynamic, value:String):Bool {
	value = value.replace('"', ''); // Strip out the quotation marks
	if (value.indexOf(":") == -1) { // If there's no colon, just check if the value exists in any field
		for (f in Reflect.fields(row)) {
			if (Std.string(Reflect.field(row, f)).toLowerCase().indexOf(value.toLowerCase()) != -1)
				return true;
		}
	} else {
		// There was a colon.  Let's split it into field:value.
		var chunks = value.split(':');
		var field = chunks.shift(); // The field is before the first colon
		var val = chunks.join(':'); // Any subsequent colons should be part of the value we look for
		if (Reflect.hasField(row, field)) { // If that field exists, check if the value is in it
			if (Std.string(Reflect.field(row, field)).toLowerCase().indexOf(val.toLowerCase()) != -1)
				return true;
		}
	}
	return false;
});

What this function does is it allows for values in the format column:value and then checks if value exists in that column. For example, a logic string we might support could be: title:harry and not "and". This filter function will resolve true for any entries where:

1. the title column contains "harry" (case-insensitive)
2. the string "and" is not in any of the columns (case-insensitive)

To see some more filtering examples, check out: docs/filters.md.

The end

That's all for now. Happy parsing!