Masonry的原理解析以及使用
Masonry应该是目前使用最为广泛的对于AutoLayout的封装(Swift版本叫做SnapKit),但是大家对于Masonry的使用只是停留在基础的方式,很少人会去理解Masonry内部去调用AutoLayout的具体原理,致使在UI上容易产生很多的冲突,导致Masonry的Crash等等情况;所以这篇文章主要是来解决上面提出的问题;
关于AutoLayout
要讲Masonry必须从iOS的布局历史开始,系统的UI布局大致分为3类:
- Frame Layout
- Auto Resizing
- Auto Layout
所谓FrameLayout即通过设置view的frame属性值从而控制view的位置以及大小;
Auto Resizing其实也是属于FrameLayout的范畴,目的就是为了让子view可以跟随superview进行大小的调整;但是不足点就是Auto Resizing无法处理同级间的view布局以及无法让superview根据子view进行反向的数据调整;
于是就出现了Autolayout,它是一种基于约束的布局系统;简单来说Autolayout的本质其实就是解析一组多元一次方程,当要确定一个视图的位置,也是需要确定视图的横纵坐标以及宽度和高度的,只是这个横纵坐标和宽度高度不再是写死的数值,而是根据约束计算得来,从而达到自动布局的效果;
约束的本质就是两个view的线性关系,上图就是一个基本的关系方程式RedView的位置其实是通过BlueView的位置来固定的;这里不做过多的讲解有兴趣的朋友可以去看官方文档
AutoLayout的使用
官方对于AutoLayout的使用提供了3种方法;但是其中两种本质都是其实都是使用NSLayoutConstraint对象进行约束;
- 使用xib以及storyboard进行布局,不过这种方式基本不用,因为不好用;
- 使用VFL语法进行约束,VFL简直就是一种又臭又长的语法,非常不好用也很难记,需要扫盲的同学可以查看官方文档,我就不多讲了,因为我自己也不是很懂;
/* Create an array of constraints using an ASCII art-like visual format string.
*/
+ (NSArray<NSLayoutConstraint *> *)constraintsWithVisualFormat:(NSString *)format options:(NSLayoutFormatOptions)opts metrics:(nullable NSDictionary<NSString *, NSNumber *> *)metrics views:(NSDictionary<NSString *, id> *)views;
- 使用NSLayoutConstraint纯代码添加,这种方式的缺点就是会大大的增加代码量,平均一个约束就需要写大量的代码,造成开发的效率大大降低;
/* Create constraints explicitly. Constraints are of the form "view1.attr1 = view2.attr2 * multiplier + constant"
If your equation does not have a second view and attribute, use nil and NSLayoutAttributeNotAnAttribute.
Use of this method is not recommended. Constraints should be created using anchor objects on views and layout guides.
*/
+ (instancetype)constraintWithItem:(id)view1 attribute:(NSLayoutAttribute)attr1 relatedBy:(NSLayoutRelation)relation toItem:(nullable id)view2 attribute:(NSLayoutAttribute)attr2 multiplier:(CGFloat)multiplier constant:(CGFloat)c;
上面曾经说过三种Autolayout的布局方法有两种的本质是一样的,就是因为VFL的本质其实就是返回多个NSLayoutConstraint对象,而不需要直接一个一个的创建NSLayoutConstraint对象大量减少代码量;个人认为要不是VFL的语法太过于变态,如果简单好用也就没有Masonry什么事了;
Masonry的本质其实就是通过链式的语法将一个一个约束关系记录下来,然后通过创建一个一个NSLayoutConstraint对象进行布局约束,Masonry内部的本质其实这样并不复杂,只是存在很多细节点,导致直接使用的它的人会存在许多疑惑点,约束间的关系理不清;
关于Masonry中的宏定义
一下有两段代码,其中一个使用了mas_定义另一个没有使用mas_定义,我想绝大部分人在使用的过程中肯定会充满疑惑,好像不管使用哪一个都没有问题;
// 使用宏定义
[view mas_makeConstraints:^(MASConstraintMaker *make) {
make.top.mas_equalTo(self.mas_top).mas_offset(20.f);
make.leading.mas_equalTo(self.mas_leading).mas_offset(20.f);
make.size.mas_equalTo(CGSizeMake(100.f, 100.f));
}];
// 不使用宏定义
[view makeConstraints:^(MASConstraintMaker *make) {
make.top.equalTo(self.top).offset(20.f);
make.leading.equalTo(self.leading).offset(20.f);
make.size.equalTo(CGSizeMake(100.f, 100.f));
}];
其实上述两种布局最后的效果都是一样的;原因就是非mas_定义的本质调用还是调用mas_定义的声明;
- (NSArray *)makeConstraints:(void(NS_NOESCAPE ^)(MASConstraintMaker *))block {
return [self mas_makeConstraints:block];
}
- (NSArray *)updateConstraints:(void(NS_NOESCAPE ^)(MASConstraintMaker *))block {
return [self mas_updateConstraints:block];
}
- (NSArray *)remakeConstraints:(void(NS_NOESCAPE ^)(MASConstraintMaker *))block {
return [self mas_remakeConstraints:block];
}
对于makeConstraints(updateConstraints && remakeConstraints)三个函数来说,本质还是调用了mas_makeConstraints(mas_updateConstraints && mas_remakeConstraints);
@property (nonatomic, strong, readonly) MASViewAttribute *mas_leading;
#define MAS_ATTR_FORWARD(attr) \
- (MASViewAttribute *)attr { \
return [self mas_##attr]; \
}
MAS_ATTR_FORWARD(leading);
对于leading这些属性来说,leading的调用实际调用的还是mas_leading;
#define mas_equalTo(...) equalTo(MASBoxValue((__VA_ARGS__)))
#define mas_greaterThanOrEqualTo(...) greaterThanOrEqualTo(MASBoxValue((__VA_ARGS__)))
#define mas_lessThanOrEqualTo(...) lessThanOrEqualTo(MASBoxValue((__VA_ARGS__)))
#define mas_offset(...) valueOffset(MASBoxValue((__VA_ARGS__)))
#ifdef MAS_SHORTHAND_GLOBALS
#define equalTo(...) mas_equalTo(__VA_ARGS__)
#define greaterThanOrEqualTo(...) mas_greaterThanOrEqualTo(__VA_ARGS__)
#define lessThanOrEqualTo(...) mas_lessThanOrEqualTo(__VA_ARGS__)
#define offset(...) mas_offset(__VA_ARGS__)
#endif
/**
* Sets the constraint relation to NSLayoutRelationEqual
* returns a block which accepts one of the following:
* MASViewAttribute, UIView, NSValue, NSArray
* see readme for more details.
*/
- (MASConstraint * (^)(id attr))equalTo;
/**
* Sets the constraint relation to NSLayoutRelationGreaterThanOrEqual
* returns a block which accepts one of the following:
* MASViewAttribute, UIView, NSValue, NSArray
* see readme for more details.
*/
- (MASConstraint * (^)(id attr))greaterThanOrEqualTo;
/**
* Sets the constraint relation to NSLayoutRelationLessThanOrEqual
* returns a block which accepts one of the following:
* MASViewAttribute, UIView, NSValue, NSArray
* see readme for more details.
*/
- (MASConstraint * (^)(id attr))lessThanOrEqualTo;
/**
* Modifies the NSLayoutConstraint constant based on a value type
*/
- (MASConstraint * (^)(NSValue *value))valueOffset;
我们可以发现对于equalTo函数默认接受的应该是一个NSObject的对象,但是我们却可以传入一个CGSize以及CGFloat这类参数,原因就是因为存在一个equalTo的宏定义调用了mas_equalTo,而mas_equalTo实际调用了equalTo函数只是将参数转换成了NSObject对象;
所以我们可以得到以下结果:
无论传入的参数是否是NSObject对象,equalTo和mas_equalTo最后调用的都是equalTo函数,需要注意的是equalTo宏定义和equalTo函数虽然本质调用一样,但是是属于不同函数;
关于Masonry的数据转换
通过上面的宏定义我们存在一个疑惑点,为一个非NSObject对象可以被equalTo接受呢?原因就是下面这个函数噶会的作用;
/**
* Given a scalar or struct value, wraps it in NSValue
* Based on EXPObjectify: https://github.com/specta/expecta
*/
static inline id _MASBoxValue(const char *type, ...) {
va_list v;
va_start(v, type);
id obj = nil;
if (strcmp(type, @encode(id)) == 0) {
id actual = va_arg(v, id);
obj = actual;
} else if (strcmp(type, @encode(CGPoint)) == 0) {
CGPoint actual = (CGPoint)va_arg(v, CGPoint);
obj = [NSValue value:&actual withObjCType:type];
} else if (strcmp(type, @encode(CGSize)) == 0) {
CGSize actual = (CGSize)va_arg(v, CGSize);
obj = [NSValue value:&actual withObjCType:type];
} else if (strcmp(type, @encode(MASEdgeInsets)) == 0) {
MASEdgeInsets actual = (MASEdgeInsets)va_arg(v, MASEdgeInsets);
obj = [NSValue value:&actual withObjCType:type];
} else if (strcmp(type, @encode(double)) == 0) {
double actual = (double)va_arg(v, double);
obj = [NSNumber numberWithDouble:actual];
} else if (strcmp(type, @encode(float)) == 0) {
float actual = (float)va_arg(v, double);
obj = [NSNumber numberWithFloat:actual];
} else if (strcmp(type, @encode(int)) == 0) {
int actual = (int)va_arg(v, int);
obj = [NSNumber numberWithInt:actual];
} else if (strcmp(type, @encode(long)) == 0) {
long actual = (long)va_arg(v, long);
obj = [NSNumber numberWithLong:actual];
} else if (strcmp(type, @encode(long long)) == 0) {
long long actual = (long long)va_arg(v, long long);
obj = [NSNumber numberWithLongLong:actual];
} else if (strcmp(type, @encode(short)) == 0) {
short actual = (short)va_arg(v, int);
obj = [NSNumber numberWithShort:actual];
} else if (strcmp(type, @encode(char)) == 0) {
char actual = (char)va_arg(v, int);
obj = [NSNumber numberWithChar:actual];
} else if (strcmp(type, @encode(bool)) == 0) {
bool actual = (bool)va_arg(v, int);
obj = [NSNumber numberWithBool:actual];
} else if (strcmp(type, @encode(unsigned char)) == 0) {
unsigned char actual = (unsigned char)va_arg(v, unsigned int);
obj = [NSNumber numberWithUnsignedChar:actual];
} else if (strcmp(type, @encode(unsigned int)) == 0) {
unsigned int actual = (unsigned int)va_arg(v, unsigned int);
obj = [NSNumber numberWithUnsignedInt:actual];
} else if (strcmp(type, @encode(unsigned long)) == 0) {
unsigned long actual = (unsigned long)va_arg(v, unsigned long);
obj = [NSNumber numberWithUnsignedLong:actual];
} else if (strcmp(type, @encode(unsigned long long)) == 0) {
unsigned long long actual = (unsigned long long)va_arg(v, unsigned long long);
obj = [NSNumber numberWithUnsignedLongLong:actual];
} else if (strcmp(type, @encode(unsigned short)) == 0) {
unsigned short actual = (unsigned short)va_arg(v, unsigned int);
obj = [NSNumber numberWithUnsignedShort:actual];
}
va_end(v);
return obj;
}
对于所有传入的参数无论是否是NSObject对象,都会通过_MASBoxValue这个函数进行数据的封装,将所有参数都转换成一个NSObject对象
- (void)setLayoutConstantWithValue:(NSValue *)value {
if ([value isKindOfClass:NSNumber.class]) {
self.offset = [(NSNumber *)value doubleValue];
} else if (strcmp(value.objCType, @encode(CGPoint)) == 0) {
CGPoint point;
[value getValue:&point];
self.centerOffset = point;
} else if (strcmp(value.objCType, @encode(CGSize)) == 0) {
CGSize size;
[value getValue:&size];
self.sizeOffset = size;
} else if (strcmp(value.objCType, @encode(MASEdgeInsets)) == 0) {
MASEdgeInsets insets;
[value getValue:&insets];
self.insets = insets;
} else {
NSAssert(NO, @"attempting to set layout constant with unsupported value: %@", value);
}
}
最后对于那些原本非NSObject对象在进行反向解析,设置对应的值;
关于Masonry的结构
下图是网上一张很详细介绍Masonry结构的一张结构图,这里引用一下,因为我不想很具体的去介绍每一句代码,只把最核心的几点告诉大家;
上图的大致流程其实很通俗易懂,因为我们这样想,我们最主要的目的无非就是讲一个一个NSLayoutConstraint约束抽象成我们能够简单通俗的编写方式,所以Masonry的主要流程其实就是每个view提供给用户一个MASConstraintMaker对象,让用户不断在MASConstraintMaker对象上添加一个一个MASConstraint的约束结构,最后将所有的MASConstraint转化成一个一个NSLayoutConstraint对象添加在相应的view上面;
接下来的很多概念都需要用到上面的结构;
关于mas_makeConstraints,mas_updateConstraints,mas_remakeConstraints的区别理解
我相信上面三个函数大家一定不会陌生,而且应该也知道对应的使用场景;
mas_makeConstraints就是创建一个新的约束
mas_updateConstraints就是更新一个原有的约束
mas_remakeConstraints就是移除现有的约束,添加新的约束;
介绍一下主要的原理,每个MASConstraintMaker对象有两个updateExisting && removeExisting属性,用来保存当前的maker的约束方式
// MASConstraintMaker
/**
* Whether or not to check for an existing constraint instead of adding constraint
*/
@property (nonatomic, assign) BOOL updateExisting;
/**
* Whether or not to remove existing constraints prior to installing
*/
@property (nonatomic, assign) BOOL removeExisting;
所以当调用mas_makeConstraints && mas_updateConstraints && mas_remakeConstraints这三个函数的时候,最后都会去执行install这个操作,而install里面本身就会判断如果是remakeConstraints那么它就会移除所有旧的约束,然后添加新的约束;对于updateConstraints && makeConstraints只是
添加新的约束,但是MASConstraint本身会保存当前的约束是更新约束还是新加约束;
// MASConstraintMaker
- (NSArray *)install {
if (self.removeExisting) {
NSArray *installedConstraints = [MASViewConstraint installedConstraintsForView:self.view];
for (MASConstraint *constraint in installedConstraints) {
[constraint uninstall];
}
}
NSArray *constraints = self.constraints.copy;
for (MASConstraint *constraint in constraints) {
constraint.updateExisting = self.updateExisting;
[constraint install];
}
[self.constraints removeAllObjects];
return constraints;
}
那么对于updateConstraints && makeConstraints内部对于updateExisting的区别其实很简单,如果updateExisting为true,那么就从当前的view去找是否存在和当前约束一样的约束,然后更新约束的constant,我们可以从layoutConstraintSimilarTo函数可以看到,判断约束是否存在的标准就是除了constant以外的所有属性;如果updateExisting为false,那么就是直接添加新的约束;
// MASConstraint
/**
* Whether or not to check for an existing constraint instead of adding constraint
*/
@property (nonatomic, assign) BOOL updateExisting;
////////////////////////////////////////////////
MASLayoutConstraint *existingConstraint = nil;
if (self.updateExisting) {
existingConstraint = [self layoutConstraintSimilarTo:layoutConstraint];
}
if (existingConstraint) {
// just update the constant
existingConstraint.constant = layoutConstraint.constant;
self.layoutConstraint = existingConstraint;
} else {
[self.installedView addConstraint:layoutConstraint];
self.layoutConstraint = layoutConstraint;
[firstLayoutItem.mas_installedConstraints addObject:self];
}
- (MASLayoutConstraint *)layoutConstraintSimilarTo:(MASLayoutConstraint *)layoutConstraint {
// check if any constraints are the same apart from the only mutable property constant
// go through constraints in reverse as we do not want to match auto-resizing or interface builder constraints
// and they are likely to be added first.
for (NSLayoutConstraint *existingConstraint in self.installedView.constraints.reverseObjectEnumerator) {
if (![existingConstraint isKindOfClass:MASLayoutConstraint.class]) continue;
if (existingConstraint.firstItem != layoutConstraint.firstItem) continue;
if (existingConstraint.secondItem != layoutConstraint.secondItem) continue;
if (existingConstraint.firstAttribute != layoutConstraint.firstAttribute) continue;
if (existingConstraint.secondAttribute != layoutConstraint.secondAttribute) continue;
if (existingConstraint.relation != layoutConstraint.relation) continue;
if (existingConstraint.multiplier != layoutConstraint.multiplier) continue;
if (existingConstraint.priority != layoutConstraint.priority) continue;
return (id)existingConstraint;
}
return nil;
}
关于Masonry为什么可以链式调用
其实Masonry可以链式调用无非就是为了缩减代码量,没有其他任何原因;从我们之前的结构图可以看到MASConstraintMaker对象包函了大量的MASConstraint属性对象,而MASConstraint属性对象里面还是定义大量的MASConstraint属性,于是就可以不断返回MASConstraint的对象;于是问题来了,它是怎么做到将每个MASConstraint对象都保存起来呢?
// MASConstraint
/**
* Usually MASConstraintMaker but could be a parent MASConstraint
*/
@property (nonatomic, weak) id<MASConstraintDelegate> delegate;
// MASConstraintMaker
@interface MASConstraintMaker () <MASConstraintDelegate>
@property (nonatomic, weak) MAS_VIEW *view;
@property (nonatomic, strong) NSMutableArray *constraints;
@end
- (MASConstraint *)constraint:(MASConstraint *)constraint addConstraintWithLayoutAttribute:(NSLayoutAttribute)layoutAttribute {
MASViewAttribute *viewAttribute = [[MASViewAttribute alloc] initWithView:self.view layoutAttribute:layoutAttribute];
MASViewConstraint *newConstraint = [[MASViewConstraint alloc] initWithFirstViewAttribute:viewAttribute];
if ([constraint isKindOfClass:MASViewConstraint.class]) {
//replace with composite constraint
NSArray *children = @[constraint, newConstraint];
MASCompositeConstraint *compositeConstraint = [[MASCompositeConstraint alloc] initWithChildren:children];
compositeConstraint.delegate = self;
[self constraint:constraint shouldBeReplacedWithConstraint:compositeConstraint];
return compositeConstraint;
}
if (!constraint) {
newConstraint.delegate = self;
[self.constraints addObject:newConstraint];
}
return newConstraint;
}
@protocol MASConstraintDelegate <NSObject>
/**
* Notifies the delegate when the constraint needs to be replaced with another constraint. For example
* A MASViewConstraint may turn into a MASCompositeConstraint when an array is passed to one of the equality blocks
*/
- (void)constraint:(MASConstraint *)constraint shouldBeReplacedWithConstraint:(MASConstraint *)replacementConstraint;
- (MASConstraint *)constraint:(MASConstraint *)constraint addConstraintWithLayoutAttribute:(NSLayoutAttribute)layoutAttribute;
@end
我们可以发现MASConstraint对象都有一个MASConstraintDelegate的代理,而MASConstraintMaker实现了这个代理,所以所有生成MASConstraint的任务其实最后都是通过MASConstraintMaker来实现的,并通过constraints进行保存;
关于Masonry如何正确添加对应的View关系
可能很多人存在一个疑惑点,Masonry是如何正确的添加每个约束关系到对应的View上呢?不理解我这句话的同学可以自行写一个约束,然后打印对应的view的constraints可以发现,每个约束都有自己的对应关系,有的添加在superview上面,有些是添加在自己的view上面;那么Masonry是怎么做的呢?
MASLayoutConstraint *layoutConstraint
= [MASLayoutConstraint constraintWithItem:firstLayoutItem
attribute:firstLayoutAttribute
relatedBy:self.layoutRelation
toItem:secondLayoutItem
attribute:secondLayoutAttribute
multiplier:self.layoutMultiplier
constant:self.layoutConstant];
layoutConstraint.priority = self.layoutPriority;
layoutConstraint.mas_key = self.mas_key;
if (self.secondViewAttribute.view) {
MAS_VIEW *closestCommonSuperview = [self.firstViewAttribute.view mas_closestCommonSuperview:self.secondViewAttribute.view];
NSAssert(closestCommonSuperview,
@"couldn't find a common superview for %@ and %@",
self.firstViewAttribute.view, self.secondViewAttribute.view);
self.installedView = closestCommonSuperview;
} else if (self.firstViewAttribute.isSizeAttribute) {
self.installedView = self.firstViewAttribute.view;
} else {
self.installedView = self.firstViewAttribute.view.superview;
}
我们发现对于存在secondView的情况,那么firstView和secondView的最近公共view就是约束需要添加的view,如果firstView是设置size的大小(包括单独的宽高),那么需要添加约束的就是自身的view,其他情况一律都是firstView的superview(比如make.center.offset(10)这类操作);
关于Masonry的一些缺省写法
很多人会在代码中会写如下的代码,以下两段代码实现效果是一样的,但是其中一段是缺省代码,那么为什么缺省的写法也是可以正确实现呢?
[view mas_makeConstraints:^(MASConstraintMaker *make) {
make.left.equalTo(self.view).offset(20.f);
make.top.equalTo(self.view).offset(20.f);
make.size.equalTo(CGSizeMake(120.f, 120.f));
}];
[view mas_makeConstraints:^(MASConstraintMaker *make) {
make.left.equalTo(self.view.left).offset(20.f);
make.top.equalTo(self.view.top).offset(20.f);
make.size.equalTo(CGSizeMake(120.f, 120.f));
}];
原因就在于Masonry会对系统的缺省值进行补充,如果在equalTo的时候传入secondViewAttribute是UIView对象,那么使用的约束类型就是该firstView的约束属性,如果传入的secondViewAttribute是secondView的约束属性,那么就直接使用;
- (void)setSecondViewAttribute:(id)secondViewAttribute {
if ([secondViewAttribute isKindOfClass:NSValue.class]) {
[self setLayoutConstantWithValue:secondViewAttribute];
} else if ([secondViewAttribute isKindOfClass:MAS_VIEW.class]) {
_secondViewAttribute = [[MASViewAttribute alloc] initWithView:secondViewAttribute layoutAttribute:self.firstViewAttribute.layoutAttribute];
} else if ([secondViewAttribute isKindOfClass:MASViewAttribute.class]) {
MASViewAttribute *attr = secondViewAttribute;
if (attr.layoutAttribute == NSLayoutAttributeNotAnAttribute) {
_secondViewAttribute = [[MASViewAttribute alloc] initWithView:attr.view item:attr.item layoutAttribute:self.firstViewAttribute.layoutAttribute];;
} else {
_secondViewAttribute = secondViewAttribute;
}
} else {
NSAssert(NO, @"attempting to add unsupported attribute: %@", secondViewAttribute);
}
}
Masonry细节解析
大家想看Masonry具体细节原理具体看这篇文章,代码上没有什么难度只要理解思想基本就能看明白;
关于Masonry的属性优先级注意点
这里要说的主要是关于View自身内容尺寸(Intrinsic Content Size),抗压缩抗拉伸(Compression-Resistance and Content-Hugging)这两种属性的概念介绍,这点是需要好好理解的,这些主要影响到的就是关于约束的优先级关系,我这里不做多讲有兴趣可以去看这篇文章;
关于Masonry的性能
其实AutoLayout的性能不是很好,想想也知道要在主线程解析多元一次方程,对于复杂的界面布局效率可想而知具体可以参见这篇文章;