Home

Awesome

Mortar

Swift 5.0 Swift 4.2 Swift 4.0

Mortar allows you to create Auto Layout constraints using concise, simple code statements.

Use this:

view1.m_right |=| view2.m_left - 12.0

Instead of:

addConstraint(NSLayoutConstraint(
    item:        view1,
    attribute:  .right,
    relatedBy:  .equal,
    toItem:      view2,
    attribute:  .left,
    multiplier:  1.0,
    constant:   -12.0
))

Other examples:

/* Set the size of three views at once */
[view1, view2, view3].m_size |=| (100, 200)

/* Pin a 200px high, same-width view at the bottom of a container */
[view.m_sides, view.m_bottom, view.m_height] |=| [container, container, 200]

/* VFL syntax */
view1 |>> viewA || viewB[==44] | 20 | viewC[~~2]

Updating from a version prior to v1.1? Read this!

A change in the default Mortar constraint priority took effect in v1.1. Please read the README_DEFAULTS.md file for more information.

Why?

Yes, there are many Auto Layout DSLs to choose from. Mortar was created to fill perceived weaknesses in other offerings:

Installing

Swift Package Manager (Preferred)

Add Mortar to your Package.swift dependency list:

.package(url: "https://github.com/jmfieldman/Mortar.git", from: "2.0.0")

Cocoapods

Notice: Cocoapods support has been deprecated; the last supported Mortar version is 1.7.0

You can install Mortar by adding it to your CocoaPods Podfile:

pod 'Mortar'

If you would like to use the Mortar VFL language:

pod 'Mortar/MortarVFL'

Or you can use a variety of ways to include the Mortar.framework file from this project into your own.

Swift Version Support

Version 2.0.0, using the Swift Package Manager, has been tested with Xcode 14.2 and Swift 5.7.

This README reflects the updated syntax and constants used in the Swift 3 Mortar release. For the Swift 2.x documentation, refer to the README_SWIFT2.md file.

pod 'Mortar', '~> 1.6'  # Swift 5.0
pod 'Mortar', '~> 1.5'  # Swift 4.2
pod 'Mortar', '~> 1.4'  # Swift 4.0
pod 'Mortar', '~> 1.3'  # Swift 3.1

Disabling MortarCreatable (Cocoapods, Legacy Swift)

The default implementation of Mortar declares a MortarCreatable protocol (create), which in legacy versions of swift causes problems with classes that do not expose the default init() method.

If you targetting a very old version of Swift, you can use:

pod 'Mortar/Core_NoCreatable'
pod 'Mortar/MortarVFL_NoCreatable'

Usage

Mortar does not require closures of any kind. The Mortar operators (|=|, |>| and |<|) instantiate and return constraints that are activated by default.

Mortar will set translatesAutoresizingMaskIntoConstraints to false for every view declared on the left side of an operator.

Equal, Less or Greater?

There are three mortar operators:

view1.m_width |=| 40                // Equal
view1.m_width |>| view2.m_width     // Greater than or equal
view1.m_size  |<| (100, 100)        // Less than or equal

Attributes

Mortar supports all of the standard layout attributes:

And iOS/tvOS spceific attributes:

It also supports composite attributes:

Implicit Attributes

Mortar will do its best to infer implicit attributes!

The m_edges attribute is implied when no attributes are declared on either side:

view1.m_edges |=| view2.m_edges     // These two lines
view1         |=| view2             // are equivalent.

If an attribute is declared on one side, it is implied on the other:

view1.m_top   |>| view2             // These two lines
view1         |>| view2.m_top       // are equivalent.

You are required to put attributes on both sides if they are not the same:

view1.m_top   |<| view2.m_bottom

Using Layout Guides

On iOS you can access the layout guides of a UIViewController. An example from inside viewDidLoad() that puts a view just below the safe top:

// Super useful when trying to position views inside a navigation/tab controller!
view1.m_top   |<| self.m_safeTop

There is also a new UIViewController property m_safeRegion to help align views to the safe region of a controller's view. To use this property you must have the MortarVFL extension installed.

// Center a view inside the safe region of a UIViewController that is a child of
// a navigation controller or tab controller
textField.m_center |=| self.m_safeRegion

Using m_safeRegion will create a "ghost" view as a subview of the controller's root view. This ghost view is hidden and non-interactive, and used only for positioning. Its class name is _MortarVFLGhostView in case you see it inside the view debugger.

Multipliers and Constants

Auto Layout constraints can have multipliers and constants applied to them. This is done with normal arithmetic operators. Attributes must be explicitly declared on the right side When arthmetic operators are used.

view1.m_size  |=| view2.m_size  * 2         // Multiplier
view1.m_left  |>| view2.m_right + 20        // Constant
view1         |<| view2.m_top   * 1.4 + 20  // Both -- m_top is implied on the left

You can also set attributes directly to constants:

view1.m_width |=| 100                // Single-dimension constants
view1.m_size  |=| 150.0              // Set multiple dimensions to the same constant
view1.m_size  |>| (200, 50)          // Set multiple dimensions to a tuple value
view1.m_frame |<| (0, 0, 50, 100)    // Four-dimension tuples supported

Arithmetic can be done using tuples for multi-dimension attributes:

view1.m_size  |=| view2.m_size + (50, 30)
view1.m_size  |>| view2.m_size * (2, 3) + (10, 10)

The special inset operator ~ operates on multi-dimension attributes:

view1         |=| view2.m_edges ~ (20, 20, 20, 20)  // view1 is inset by 20 points on each side

Attributes in Tuples

You are allowed to put attributes inside of tuples:

view1.m_size  |=| (view2.m_width, 100)

Multiple Simultaneous Constraints

Multiple constraints can be created using arrays:

view1 |=| [view2.m_top, view3.m_bottom, view4.m_size]

/* Is equivalent to: */
view1 |=| view2.m_top
view1 |=| view3.m_bottom
view1 |=| view4.m_size
[view1, view2, view3].m_size |=| (100, 200)

/* Is equivalent to: */
[view1.m_size, view2.m_size, view3.m_size] |=| (100, 200)

/* Is equivalent to: */
view1.m_size |=| (100, 200)
view2.m_size |=| (100, 200)
view3.m_size |=| (100, 200)

This might be a convenient way to align an array of views, for example:

[view1, view2, view3].m_centerY |=| otherView
[view1, view2, view3].m_height  |=| otherView

If you put arrays on both sides of the constraint, it will only constrain elements at the same index. That is:

[view1.m_left, view2, view3] |=| [view4.m_right, view5, view6]

/* Is equivalent to: */
view1.m_left |=| view4.m_right
view2        |=| view5
view3        |=| view6

You can use this to create complex constraints on one line. For example, to create a 200-point high view that sits at the bottom of a container view:

[view.m_sides, view.m_bottom, view.m_height] |=| [container, container, 200]

Priority

You can assign priority to constraints using the ! operator. Valid priorities are:

v0 |=| self.container.m_height
v1 |=| self.container.m_height ! .low
v2 |=| self.container.m_height ! .medium
v3 |=| self.container.m_height ! .high
v4 |=| self.container.m_height ! .required
v5 |=| self.container.m_height ! 300

You can also put priorities inside tuples or arrays:

view1        |=| [view2.m_caps   ! .high, view2.m_sides      ! .low]  // Inside array
view1.m_size |=| (view2.m_height ! .high, view2.m_width + 20 ! .low)  // Inside tuple

Default Priority

Defaults have changed in Mortar v1.1; See README_DEFAULTS.md if you are updating.

By default, constraints are given priority of .required which is equal to 1000 (out of 1000) and is the same default used by Apple's constraint methods. Sometimes you may want large batches of constraints to have a different priority, and it is messy to include something like ! .medium after every constraint.

You can change the global base default value by using set:

MortarDefault.priority.set(base: .medium)

You can use this in the AppDelegate to change the app-wide default constraint priority.

Because this can only be changed on the main thread, it is safe to call just before your layout code. Keep in mind it will affect all future Mortar contraints! If you are adjusting the default for a single layout section, it is usually wiser to use the stack mechanism to change the default priority used in a frame of code:

MortarDefault.priority.push(.low)

v1 |=| v2 // Given priority .low automatically
...

MortarDefault.priority.pop()

You may only call the push/pop methods on the main thread, and Mortar will raise an exception if you do not properly balance your pushes and pops.

Change Priority

You can change the priority of a MortarConstraint or MortarGroup by calling the changePriority method. This takes either a MortarLayoutPriority enum, or a UILayoutPriority value:

let c = view1 |=| view2 ! .low     // Creates 4 low-priority constraints (1 per edge)
c.changePriority(to: .high)        // Sets all 4 constraints to high priority

Remember that you can't switch to or from Required from any other priority level (this is an Auto Layout limitation.)

Create Deactivated Constraints

You can use the ~~ operator as a shorthand for constraint activation and deactivation. This makes the most sense as part of constraint declarations when you want to create initially-deactivated constraints:

let constraint = view1 |=| view2 ~~ .deactivated

// Later on, it makes more semantic sense to call .activate():
constraint.activate()

// Even though this is functionally equivalent:
constraint ~~ .activated

// It works with groups too:
let group = [
    view1 |=| view2
    view3 |=| view4
] ~~ .deactivated

Keeping Constraint References

The basic building block is the MortarConstraint, which wraps several NSLayoutConstraint instances that are relevant to multi-affinity attributes like m_frame (4) or m_size (2).

You can capture a MortarConstraint for later reference:

let constraint = view1.m_top |=| view2.m_bottom

The raw NSLayoutConstraint elements can be accessed through the layoutConstraints accessor:

let mortarConstraint = view1.m_top |=| view2.m_bottom
for rawLayoutConstraint in mortarConstraint.layoutConstraints {
    ...
}

You can create an entire group of constraints:

let group = [
    view1.m_origin |=| view2,
    view1.m_size   |=| (100, 100)
]

Mortar includes a convenient typealias to refer to arrays of MortarConstraint objects:

public typealias MortarGroup = [MortarConstraint]

You can now activate/deactivate constraints:

let constraint = view1.m_top |=| view2.m_bottom

constraint.activate()
constraint.deactivate()

let group = [
    view1.m_origin |=| view2,
    view1.m_size   |=| (100, 100)
]

group.activate()
group.deactivate()

Replacing Constraints and Groups of Constraints

Constraints and groups have a replace method that deactives the target and activates the parameter:

let constraint1 = view1.m_sides |=| view2
let constraint2 = view1.m_width |=| view2 ~~ .deactivated

constraint1.replace(with: constraint2)

let group1 = [
    view1.m_sides |<| view2,
    view1.m_caps  |>| view2,
]

let group2 = [
    view1.m_width |=| view2
] ~~ .deactivated

group1.replace(with: group2)

Compression Resistance and Content Hugging

Mortar provides some shorthand properties to adjust a view's compression resistance and content hugging priorities:

// Set both horizontal and vertical compression resistance priority simultaneously:
view1.m_compResist = 1

// Set horizontal and vertical compression resistance independently:
view1.m_compResistH = 300
view1.m_compResistV = 800

// Set both horizontal and vertical content hugging priority simultaneously:
view1.m_hugging = 1

// Set horizontal and vertical content hugging independently:
view1.m_huggingH = 300
view1.m_huggingV = 800

You can get the horizontal and vertical values independently, but not together:

// These getters are fine:
let c1 = view1.m_compResistH
let c2 = view1.m_compResistV
let h1 = view1.m_huggingH
let h2 = view1.m_huggingV

// These getters raise exceptions:
let cr = view1.m_compResist
let hg = view1.m_hugging

MortarVFL

Mortar supports a VFL language that is roughly equivalent to Apple's own Auto Layout VFL langauge. The primary advanges are:

MortarVFL is contrained to its own extension because it makes heavy use of custom operators. These operators may not be compatible with other libraries you are using, so we don't want Mortar core to conflict with those.

pod 'Mortar/MortarVFL'

MortarVFL Internal Composition

The heart of a MortarVFL statement is a list of VFL nodes that are positioned sequentially along either the horizontal or vertical axis. A node list might look like:

viewA | viewB[==viewA] || viewC[==40] | 30 | viewD[~~1] | ~~1 | viewE[~~2]

// viewA has a size determinde by its intrinsic content size
// viewA is separated from viewB by 0 points (| operator)
// viewB has a size equal to viewA
// viewB is separated from viewC by the default padding (8 points; || operator)
// viewC has a fixed size of 40
// viewC is separated from viewD by a space of 30 points
// viewD has a weighted size of 1
// viewD is separated fom viewE by a weighted space of 1
// viewE has a weighted size of 2

VFL nodes:

Nodes are separated by either a | or || operator.

The | operator introduces zero extra distance between nodes. You can use this operator to connect nodes directly with zero spacing, or insert your own fixed/weighted numerical value between them (e.g. | 30 | or | ~~2 |). In these cases, the 30 and ~~2 are considered nodes that represent whitespace (no attached view).

The || operator separates nodes by the default padding (8 points).

Nodes that represent views respect their intrinsic content as much as possible given the realvent constraints and priorities. View nodes can also constain a subscript that gives them a size constraint. You can use [==#] to give the view a fixed size, or [~~#] to give the view a weighted size. You can also reference other views, e.g. [==viewA] to give the node's view the same constraint as the one it references.

MortarVFL will throw an error if you have cyclic view references, e.g. viewA[==viewB] | viewB[==viewA]

Arrays in a Node

As an advanced technique, you can use an array of views in a node. It would look something like this:

viewA || [viewB, viewC, viewD][==40] || viewE

This positions the arrayed nodes in parallel with each other. In the above example, all three of viewB, viewC and viewD will be sized 40 points and be adjacent to viewA and viewE. This is very useful for complex grid-based layouts.

Capture

MortarVFL statements must be captured on at least one end by a view attribute. These captures look something like:

// viewB and viewC will take equal width between the
// right edge of viewA and the left edge of viewD
viewA.m_right |> viewB[~~1] | viewC[~~1] <| viewD.m_left

// viewB and viewC will be equal width between the
// left/right edges of viewA, inset by 8pt padding
// and separated by 40pts.
viewA ||>> viewB[~~1] | 40 | viewC[~~1]

MortarVFL support horizontal and vertical spacing in a similar manner. The horizontal operators use the > character while the vertical operators use the ^ character. Otherwise they act similarly. For example, the vertical version of the above statement would be:

// viewB and viewC will take equal height between the
// bottom edge of viewA and the top edge of viewD
viewA.m_bottom |^ viewB[~~1] | viewC[~~1] ^| viewD.m_top

Mortar will make sure your operators are compatible with the attributes you've selected. For example, using |> with m_top would be an axis mismatch and raise an exception.

Implicit Capture Attributes

If you don't provide attributes on the capture terminals, Mortar will derive them based on the axis and position:

// These are equivalent:
viewA.m_left |> viewB | viewC <| viewD.m_right
viewA        |> viewB | viewC <| viewD

// These are equivalent:
viewA.m_top  |^ viewB | viewC ^| viewD.m_bottom
viewA        |^ viewB | viewC ^| viewD

Important Observation: Implicit attributes might be opposite of what you expect. This is because implicit attributes are normally used to capture views inside the bounds of parent views and so we use the outer edges, not the inner edges.

Implicit Surround

If you want the MortarVFL nodes to be inside the bounds of a single view, you can use the surround operators instead of placing the same view at both terminals.

The surround operators use either >> or ^^:

// viewB and viewc will be equal width between the
// left/right edges of viewA, inset by 8pt padding
// and separated by 40pts.
viewA ||>> viewB[~~1] | 40 | viewC[~~1]

// viewB will be twice as tall as viewC; both will be between
// the top/bottom edges of viewA.
viewA |^^ viewB[~~2] | viewC[~~1]

// Using m_visibleRegion is helpful for layouts in child view controllers
// to get views laid out inside the visible region, not under nav/tab bars
self.m_visibleRegion ||^^ viewA | viewB | viewC

Single-Ended Statements

Up until now, all of the examples have shown statements bordered by two attributes (left and right, top and bottom).

For statements surrounded on both sides, you cannot have all fixed spacing. This means you will need at least one weighted or intrinsically sized node. This allows Mortar to make your constraints flexible between the terminals. You may see odd behavior if you only have intrinsically sized nodes, and their compression resistance and content hugging are artificially forced to .required.

For statements that have a single terminal, the opposite is true. You cannot use any weight-based nodes, and they must all be fixed size or intrinsic content size. This is because there is no second endpoing to use as an anchor for relative sizing.

Single-terminal statements look the same as the others, but trailing operators use a bang: ! Unfortunately this looks very much like the pipe operator, so don't be confused. Specifically, when just attaching a statement to a trailing attribute, use <!, <!!, ^! or ^!!.

// viewB will be placed at the right edge of viewA and be 44pts wide.
// viewC will be placed 8pts (padding) right of viewB and will be 88pts wide.
viewA.m_right |> viewB[==44] || viewC[==88]

// viewC will be placed at the left edge of viewA and be 88pts wide.
// viewB will be placed 8pts (padding) left of viewC and will be 44pts wide.
viewB[==44] || viewC[==88] <! viewA.m_left

// viewB will be placed at the bottom edge of viewA and be 44pts high.
// viewC will be placed 8pts below viewB and respect its intrinsic content height.
viewA.m_bottom |^ viewB[==44] || viewC

// viewC will be placed at the top edge of viewA and be 88pts high.
// viewB will be placed 8pts above viewC and will be 44pts high.
viewB[==44] || viewC[==88] ^! viewA.m_top

Again, note the use of the ! bang symbol for trailing single-ended statements, and that there are no weight-based nodes. Leading single-ended statements use the operator with the pipe: |>

Examples

There are several examples of MortarVFL in the Examples/MortarVFL project.

Visual View Hierarchy Creation

Mortar provides the |+| and |^| operators to quickly add a subview or array of subviews. This can be used to create visual expressions of the view hierarchy.

Now this:

self.view.addSubview(backgroundView)
self.view.addSubview(myCoolPanel)
myCoolPanel.addSubview(nameLabel)
myCoolPanel.addSubview(nameField)

Turns into:

    self.view |+| [
        backgroundView,
        myCoolPanel |+| [
            nameLabel,
            nameField
        ]
    ]

Alternatively, if you want to see the upper subviews at the beginning of the array (so that visually, the views closer to the top of the file are closer to the user), use the |^| operator:

    self.view |^| [
        myCoolPanel |^| [      // myCoolPanel is added second and
            nameField,         // is therefore on top of backgroundView
            nameLabel
        ],
        backgroundView
    ]

Initializing NSObject at Creation

Mortar extends NSObject with the create class function. This class function performs a parameter-less instantiation of the class, and passes the new instance into the provided closure. This allows you to configure an instance at creation-time, which is really nice for compartmentalizing view configuration.

As you can see in the below example, configuration of the view is separated from the code needed to attach it to the view controller hierarchy and layout.

class MyController: UIViewController {

    // Instantiation/configuration
    let myLabel = UILabel.create {
        $0.text          = "Some Text"
        $0.textAlignment = .center
        $0.textColor     = .red
    }

    // UIViews can use the immediate init block
    let myButton = UIButton {
        $0.setTitle("Hello", for: .normal)
    }

    override func viewDidLoad() {
        super.viewDidLoad()

        // Hierarchy
        self.view |+| [
            myLabel,
            myButton
        ]

        // Layout
        myLabel.m_top     |=| self.view
        myLabel.m_centerX |=| self.view
        myButton.m_width  |=| myLabel
    }
}

Combining Hierarchy and Layout

One of the historical problems with combining hierarchy and layout into a single function (like SwiftUI) is that UIKit requires two views to be in the same view hierarchy before a constrait is activated.

This prevented you from doing something like:

view1 |+| [
    UILabel.create {
        // Crash here, because the newly created UILabel is not
        // a subview of view1 until *after* the top-level |+| is
        // executed.
        $0.m_width |=| view1
    }
]

As of version 2.0.0, Mortar understands how to defer constraint activation while the hierarchy is being created:

view1 |+| [
    UILabel.create {
        // This is OK in Mortar 2.0.0; the constraint will not
        // activate until after the outer-most |+| executes.
        $0.m_width |=| view1
    }
]

That is, any time the |+| or |^| operators are in progress, or you use the new .addSubviews or .addArrangedSubviews result builders, Mortar knows not to activate any constraint assignments until after the outer-most hierarchy assignment is completed.

Mortar 2.0.0 now also offers result builders as the right hand side of the add subview operators. The result builders take the left-side view in as a block parameter, which lets children easily reference anonymous parents:

view |+| { view in
    UIStackView {
        $0.spacing = 1
        $0.axis = .vertical
        $0.m_width |=| view
    } |+| { stack in
        UILabel {
            $0.text = viewModel.helloText
            $0.m_height |=| stack
        }
        UILabel {
            $0.text = "World"
            $0.m_height |=| stack
            $0.reactive.text <~ viewModel.worldText
        }
        UIButton {
            $0.reactive.pressed = viewModel.pressed
        }
    }
]

Combining this with a reactive framework like ReactiveSwift can get you nearly all the way to a single-expression view definition, even for large custom layouts that can leverage stack views properly.