Inquire
Inquire allows you to generate advanced query strings. Currently supports armet syntax query strings.
Table of Contents
- Installation
- The Problem
- A Solution
- Background
- Usage
- Examples
Installation
Inquire is available on the npm registry: inquire
You can install it with npm.
npm install inquireThe Problem
Currently, query strings only conjoin predicates together with equality. Armet attempts to extend this in two ways:
- Allowing predicates to be conjoined, disjoined and negated.
- Allowing more than just equality, e.g. inequalities.
This ends up creating a problem. We now have to generate these advanced query strings. Generating these advanced query strings by hand gets unwieldy.
For example, let's say we are dealing with a REST api for shapes.
We can GET on /api/shape and return all of the shapes.
Now let's try to get more detailed in our result.
Let's say we want red shapes wider than 30 pixels or
we want all shapes with no more than 12 sides or
we want all of the squares that either aren't black or that were created by bob.
Our GET request becomes:
/api/shape?(color=red&width>30);sides<=12;(shape=square&(color!=black;user=bob))
A Solution
We need a better way to generate query strings for armet's consumption. Inquire allows you do the same query without trying to man-handle the string.
import Inquire
query = (("color" `eq` "red") |&| ("width" `gt` "30")) ||| ("sides" `le` "12") ||| (("shape" `eq` "square") |&| (("color" `ne` "black") ||| ("user" `eq` "bob")))
url = "/api/shape?" ++ show query
-- url => "/api/shape?(color=red&width>30);sides<=12;(shape=square&(color!=black;user=bob))"LiveScript:
require! I: \inquire.Inquire
query = I.or(I.or(I.and(I.eq(\color)(\red))(I.gt(\width)(30)))(I.le(\sides)(12)))(I.and(I.eq(\shape)(\square))(I.or(I.ne(\color)(\black))(I.eq(\user)(\bob))))
url = "/api/shape?#{I.generate query}"
# url => '/api/shape?(color=red&width>30);sides<=12;(shape=square&(color!=black;user=bob))'Javascript:
var I = require('inquire').Inquire;
var query = I.or(I.or(I.and(I.eq('color')('red'))(I.gt('width')(30)))(I.le('sides')(12)))(I.and(I.eq('shape')('square'))(I.or(I.ne('color')('black'))(I.eq('user')('bob'))))
var url = '/api/shape?' + I.generate(query)
// url => '/api/shape?(color=red&width>30);sides<=12;(shape=square&(color!=black;user=bob))'Note: Inquire removes as many parens as logically possible.
Of course this still looks horrible to have to deal with because it's all on one line. So, as with most software issues, it's better to decompose this into smaller pieces and then rejoin the pieces at the end.
Background
N.B. Please report any misinformation, pull requests are welcome!
Basics
At its core Inquire is a Boolean Algebra. This means that all of the properties of other boolean algebras can be applied to Inquire. This includes things like:
- Associativity
- Commutativity
- Idempotence
- DeMorgan's Laws
- Implication
- Equivalence
Because of this, we can optimize our queries. In some cases, we can optimize away the entire query!
The basic building block of Inquire is a predicate.
Each Predicate has a key, a relational operator and a value.
The supported relational operators are ==, !=, <, <=, >, and >=.
A key or a value can be any thing.
N.B. Currently in JS, only the basic JS types are supported (Boolean, Number, String).
We can construct more complex queries by joining our predicates together.
There are two binary combinators and a unary combinator: and, or and not, respectively.
Deeper
Also, since Boolean Algebras are Complemented Distributive Lattices, we get all of the properties of those for free as well!
The complemented part means that every query has a complement. We construct the complement for an entire query by wrapping it in a negation. This is the lazy way, and not actually true to the definition. The way to actually get the complement is to take each predicate and change its relational operator.
== |
<=> | != |
> |
<=> | <= |
< |
<=> | >= |
The distributive part means that our binary combinators can be distributed over each other. This is similar to how multiplication can distribute over addition in Semirings like the Natural numbers with 0, or Integers or the reals, etc. The difference is, each operation can distribute over the other.
So the following are equivalent: p AND (q OR r), (p OR q) AND (p OR r)
As well as: p OR (q AND r), (p AND q) OR (p AND r)
The lattice part has two equivalent interpretations. One states that there is a poset nestled underneath our structure. The other states that we have some of the axioms mentioned above.
Usage
Inquire has a few basic combinators on which most of everything else is built. Since Inquire is tied closely to armet, most of this is about filtering responses based on model attributes. However, the fundamentals are the same for any purpose.
Predicate combinators
eq
k -> v -> Inquire k v
Constructs an Inquire from two values with the relation ==.
Will return results where the key on the model equals the supplied val
ne
k -> v -> Inquire k v
Constructs an Inquire from two values with the relation !=
Will return results where the key on the model does not equal the supplied val
gt
k -> v -> Inquire k v
Constructs an Inquire from two values with the relation >
Will return results where the key on the model is greater than the supplied val
ge
k -> v -> Inquire k v
Constructs an Inquire from two values with the relation >=
Will return results where the key on the model is greater than or equal to the supplied val
lt
k -> v -> Inquire k v
Constructs an Inquire from two values with the relation <
Will return results where the key on the model is less than the supplied val
le
k -> v -> Inquire k v
Constructs an Inquire from two values with the relation <=
Will return results where the key on the model is less than or equal to the supplied val
Junction combinators
and
Inquire k v -> Inquire k v -> Inquire k v
Constructs an Inquire from two Inquires with the conjunction &
Will return results where both left and right subqueries are true.
or
Inquire k v -> Inquire k v -> Inquire k v
Constructs an Inquire from two Inquires with the disjunction ;
Will return results where either one, or both, left and right subqueries are true.
Wrap combinator
not
Inquire k v -> Inquire k v
Constructs an Inquire from one Inquire with the negation !
Will return results where subquery is false.
Derived combinators
From these simple combinators, we can create more complex ideas:
implies
Inquire k v -> Inquire k v -> Inquire k v
Constructs an Inquire from two Inquires.
This is the direct definition of implication, namely (NOT p) OR q.
equiv
Inquire k v -> Inquire k v -> Inquire k v
Constructs an Inquire from two Inquires.
This is the direct definition of equivalence, namely (p AND q) OR ((NOT p) AND (NOT q)).
xor
Inquire k v -> Inquire k v -> Inquire k v
Constructs an Inquire from two Inquires.
This is the direct definition of exclusive or, namely (p AND (NOT q)) OR ((NOT p) AND q).
absorb
Inquire k v -> Inquire k v
Simplifies specific Inquires.
p AND (p OR q) => p
p OR (p AND q) => p
associate
Inquire k v -> Inquire k v
Reorders parens from what they are now to the other possible way.
p AND (q AND r) => (p AND q) AND r
(p AND q) AND r => p AND (q AND r)
p OR (q OR r) => (p OR q) OR r
(p OR q) OR r => p OR (q OR r)
assocLeft
Inquire k v-> Inquire k v
Reorders parens to the left.
p AND (q AND r) => (p AND q) AND r
p OR (q OR r) => (p OR q) OR r
assocRight
Inquire k v-> Inquire k v
Reorders parens to the right.
(p AND q) AND r => p AND (q AND r)
(p OR q) OR r => p OR (q OR r)
commute
Inquire k v -> Inquire k v
Reorders queries.
p AND q => q AND p
p OR q => q OR p
distribute
Inquire k v -> Inquire k v
Distributes one query over its dual.
p AND (q OR r) => (p OR q) AND (p OR r)
p OR (q AND r) => (p AND q) OR (p AND r)
codistribute
Inquire k v -> Inquire k v
Performs the opposite of distribution.
(p OR q) AND (p OR r) => p AND (q OR r)
(p AND q) OR (p AND r) => p OR (q AND r)
idempotent
Inquire k v -> Inquire k v
Simplifies the Inquire with the Idempotency rule.
p AND p => p
p OR p => p
Construction
Creating new Inquires is done with either single key val arguments,
an array of [key, val] pairs,
or an array of {key: k, val: v} objects.
The predicate combinators above correspond to the single key, val option.
fromArray
[Inquire k v] -> Inquire k v
This constructs a single Inquire from the conjunction of all the Inquires in the array.
fromArrayPair
[[k, v]] -> Inquire k v
This constructs an Inquire with the default relation of == for all predicates,
and the default junction of & for all combinations of predicates.
fromArrayObj
[{key: k, val: v}] -> Inquire k v
This constructs an Inquire with the default relation of == for all predicates,
and the default junction of & for all combinations of predicates.
This expects each object to have two entries: key and val.
These should correspond to the intended predicate key and val.
This constructs an Inquire with the default relation of == for all predicates,
and the default junction of & for all combinations of predicates.
Modification
Since Inquire is also a Foldable and Functor,
it provides the ability to fold and map over all of vals, respectively.
These operations are similar to the native reduce and map for javascript arrays,
the fantasy land(fantasy land) specification,
or something you might get from a library like underscore.
So, if you can use Array.map, _.reduceRight or similar,
you already know how to use these functions.
foldr
(v -> v' -> v') -> v' -> Inquire k v -> v'
This is a right associative "catamorphism". What that means is that it takes each value and joins it to the next given the supplied function.
The first argument is a binary function. The second argument is the initial value. The last argument is the Inquire.
foldl
(v' -> v -> v') -> v' -> Inquire k v -> v'
This is a left associative "catamorphism". What that means is that it takes each value and joins it to the next given the supplied function.
The first argument is a binary function. The second argument is the initial value. The last argument is the Inquire.
map
(v -> v') -> Inquire k v -> Inquire k v'
This will modify every value in the Inquire with the given function.
There are a few modification combinators available.
filterByKey
(k -> Boolean) -> Inquire k v -> Inquire k v
Removes all predicates that fail the supplied boolean test.
The first argument must be a function that tests a key and returns a boolean.
filterByVal
(v -> Boolean) -> Inquire k v -> Inquire k v
Removes all predicates that fail the supplied boolean test.
The first argument must be a function that tests a val and returns a boolean.
findByKey
k -> Inquire k v -> Maybe (Inquire k v)
Searches for the first element which has the specified key.
If successful, will return a Just predicate.
If unsuccessful, will return a Nothing.
This is biased to the right.
findByVal
v -> Inquire k v -> Maybe (Inquire k v)
Searches for the first element which has the specified val.
If successful, will return a Just predicate.
If unsuccessful, will return a Nothing.
This is biased to the right.
remo
:: (Inquire k v -> Inquire k v -> Boolean) -> Inquire k v -> Inquire k v -> Inquire k v
Helper function for remove and removeAll.
Requires a binary function that takes two inquires and returns a boolean indicating whether the to continue removal.
remove
Inquire k v -> Inquire k v -> Inquire k v
Removes the rightmost instance of the first Inquire from the second Inquire.
removeAll
Inquire k v -> Inquire k v -> Inquire k v
Removes every instance of the first Inquire from the second Inquire.
replaceValByKey
v -> k -> Inquire k v -> Inquire k v
Replaces each Inquire that has the specified key with the given val.
The first argument is the val to replace with.
The second argument is the key to test with.
replaceValByVal
v -> v -> Inquire k v -> Inquire k v
Replaces each Inquire that has the specified val with the given val.
The first argument is the val to replace with.
The second argument is the val to test with.
unsafeFindByKey
k -> Inquire k v -> Maybe (Inquire k v)
Unsafe version of findByKey.
This is unsafe in the sense that it uses Prelude.unsafeRefEq for equality testing.
AKA === in javascript.
So it could fail depending on what the values are. Use with caution!
However, it provides a nice wrapper around the safe version. i.e. you don't have to pass a two typeclass dictionary references to the safe version.
unsafeFindByVal
v -> Inquire k v -> Maybe (Inquire k v)
Unsafe version of findByVal.
This is unsafe in the sense that it uses Prelude.unsafeRefEq for equality testing.
AKA === in javascript.
So it could fail depending on what the values are. Use with caution!
However, it provides a nice wrapper around the safe version. i.e. you don't have to pass a two typeclass dictionary references to the safe version.
unsafeFromObj
{ | a } -> Inquire k v
Unsafe way to drop an object into an Inquire.
This just iterates through the object and tries to take all the top level keys and vals as arguments to the Inquire.
Conjoins everything with the relation equals.
Note, there's no telling what will happen if you put in something that's not a simple key val object.
unsafeRemove
Inquire k v -> Inquire k v -> Inquire k v
Unsafe version of remove.
This is unsafe in the sense that it uses Prelude.unsafeRefEq for equality testing.
AKA === in javascript.
So it could fail depending on what the values are. Use with caution!
However, it provides a nice wrapper around the safe version. i.e. you don't have to pass a two typeclass dictionary references to the safe version.
unsafeRemoveAll
Inquire k v -> Inquire k v -> Inquire k v
Unsafe version of removeAll.
This is unsafe in the sense that it uses Prelude.unsafeRefEq for equality testing.
AKA === in javascript.
So it could fail depending on what the values are. Use with caution!
However, it provides a nice wrapper around the safe version. i.e. you don't have to pass a two typeclass dictionary references to the safe version.
unsafeReplaceValByKey
v -> k -> Inquire k v -> Inquire k v
Unsafe version of replaceValByKey.
This is unsafe in the sense that it uses Prelude.unsafeRefEq for equality testing.
AKA === in javascript.
So it could fail depending on what the values are. Use with caution!
However, it provides a nice wrapper around the safe version. i.e. you don't have to pass a two typeclass dictionary references to the safe version.
unsafeReplaceValByVal
v -> v -> Inquire k v -> Inquire k v
Unsafe version of replaceValByVal.
This is unsafe in the sense that it uses Prelude.unsafeRefEq for equality testing.
AKA === in javascript.
So it could fail depending on what the values are. Use with caution!
However, it provides a nice wrapper around the safe version. i.e. you don't have to pass a two typeclass dictionary references to the safe version.
unsafeIsInquire
a -> Boolean
This is unsafe in many senses. Primarily it relies on the current implementation of the compiler to check if any js value is an Inquire. Use with caution as it can generate false positives.
Zipper
The Zipper is a purely functional data structure for manipulating other data structures.
An immense amount of information can be found on Zippers.
It was first written about as a Functional Pearl by Gérard Huet.
The basic idea is that you create a hole in a data structure that you want to modify,
and keep track of the context to that hole.
Once you modify the hole, you can then use the context to find your way back to the top of the structure.
In our case, the hole will be an Inquire, and the context the path to the top of the Inquire.
We can move in the four cardinal directions: Up, Down, Left, Right.
This allows us to go to any position in our Inquire,
modify that Predicate, Junction, or Wrap, and return the new structure easily.
toInquireZ
Inquire k v -> InquireZ k v
Creates a new Zipper from an Inquire. This is how we can start to manipulate our Inquire.
fromInquireZ
InquireZ k v -> Inquire k v
Returns our modified Inquire from a Zipper.
zipLeft
InquireZ k v -> Maybe (InquireZ k v)
Moves to the left of the current hole, if possible, or does nothing.
zipRight
InquireZ k v -> Maybe (InquireZ k v)
Moves to the right of the current hole, if possible, or does nothing.
zipDown
InquireZ k v -> Maybe (InquireZ k v)
Moves down or to the right of the current hole, if possible, or does nothing.
zipUp
InquireZ k v -> Maybe (InquireZ k v)
Moves up of the current hole, if possible, or does nothing.
zipUpmost
InquireZ k v -> InquireZ k v
Moves to the top of the Zipper.
zipLeftmost
InquireZ k v -> InquireZ k v
Moves as far to the left of the Zipper as possible.
zipRightmost
InquireZ k v -> InquireZ k v
Moves as far to the right of the Zipper as possible.
getHole
InquireZ k v -> Inquire k v
Peeks into the Zipper and returns the current Inquire in the hole.
query
(Inquire k v -> a) -> InquireZ k v -> a
Applies a function to the current Inquire in the hole and returns the result.
modify
(Inquire k v -> Inquire k v) -> InquireZ k v -> InquireZ k v
Applies a function to the current Inquire in the hole.
Examples
So let's say we've got a Todo list available at some endpoint. We'd like to retrieve all of the items in the list that are due after tax day 2014. We'll use moment to make things a bit easier.
> inquire = require('inquire')
> I = inquire.Inquire
> moment = require('moment')
> query0 = I.gt('due')(moment('Apr 15, 2014'))
> I.generate(query0)
'due>Tue%20Apr%2015%202014%2000%3A00%3A00%20GMT-0700'Later on if you decide to change the date value, there are combinators that make it much easier than doing this by hand:
> IC = inquire.Inquire_Combinators
> query1 = IC.unsafeReplaceValByKey('tonight')('due')(query0)
> I.generate(query1)
'due>tonight'Maybe instead you want to move the cut-off point a month sooner, so March 15, 2014. This can be done using the fact that Inquire is a Functor.
> prevMonth = function(x) { return moment(x).subtract('month', 1); }
> query2 = IC.map(prevMonth)(query0)
> I.generate(query2)
'due>Sat%20Mar%2015%202014%2000%3A00%3A00%20GMT-0700'Of course, this will fail miserably if you have more values that aren't moments. The more robust way currently is to use the Zipper. Let's start by making our query a bit bigger and seeing it fail.
> query3 = I.and(query0)(IC.fromArrayPair([['user', 'Joe'], ['subject', 'bills']]))
> I.generate(query3)
'due>Tue%20Apr%2015%202014%2000%3A00%3A00%20GMT-0700&subject=bills&user=Joe'
> query4 = IC.map(prevMonth)(query3)
> I.generate(query4)
'due>Sat%20Mar%2015%202014%2000%3A00%3A00%20GMT-0700&subject=Invalid%20date&user=Invalid%20date'So as we can see, it failed because the function was expecting to operate on moments.
N.B. This wouldn't be a problem in a language like purescript, as the line wouldn't have even compile. This is only a concern in javascript because it does not respect types.
So let's use our zipper to find that specific date, and change it. We know it's the leftmost predicate in our query, so let's go there, and modify that.
The basic idea with the Zipper is that it allows you to traverse a data structure, modify some part of it, and then return the new modified structure.
> IZ = inquire.Inquire_Zipper
> zip0 = IZ.toInquireZ(query3)
> zip1 = IZ.zipLeftmost(zip0)
> zip2 = IZ.modify(IC.map(prevMonth))(zip1)
> query5 = IZ.fromInquireZ(zip2)
> I.generate(query5)
'due>Sat%20Mar%2015%202014%2000%3A00%3A00%20GMT-0700&subject=bills&user=Joe'Wonderful! We see that the Zipper made it so our logic was expressed almost verbatim in the code. It was composable and reusable.