model定义时,每个field都是一个类属性,一个对象。在生成类时,属性有contribute_to_class的方法,会调用该方法。

m2m field,它会先调用自己的contribute_to_class,然后调用父类的contribute_to_class,因为m2m类是继承自relatfield,

def contribute_to_class(self, cls, name, virtual_only=False):

        super(RelatedField, self).contribute_to_class(cls, name, virtual_only=virtual_only)#在option添加一个field

        self.opts = cls._meta

        if not cls._meta.abstract:
if self.remote_field.related_name:
related_name = force_text(self.remote_field.related_name) % {
'class': cls.__name__.lower(),
'app_label': cls._meta.app_label.lower()
}
self.remote_field.related_name = related_name def resolve_related_class(model, related, field):
field.remote_field.model = related
field.do_related_class(related, model)
lazy_related_operation(resolve_related_class, cls, self.remote_field.model, field=self)
def do_related_class(self, other, cls):#关键是调用这个函数时把两个model类调换了,model与related位置对调。
self.set_attributes_from_rel()
self.contribute_to_related_class(other, self.remote_field) def lazy_related_operation(function, model, *related_models, **kwargs):
"""
Schedule `function` to be called once `model` and all `related_models`
have been imported and registered with the app registry. `function` will
be called with the newly-loaded model classes as its positional arguments,
plus any optional keyword arguments. The `model` argument must be a model class. Each subsequent positional
argument is another model, or a reference to another model - see
`resolve_relation()` for the various forms these may take. Any relative
references will be resolved relative to `model`. This is a convenience wrapper for `Apps.lazy_model_operation` - the app
registry model used is the one found in `model._meta.apps`.
"""
models = [model] + [resolve_relation(model, rel) for rel in related_models]
model_keys = (make_model_tuple(m) for m in models)
apps = model._meta.apps
return apps.lazy_model_operation(partial(function, **kwargs), *model_keys) def lazy_model_operation(self, function, *model_keys):
"""
Take a function and a number of ("app_label", "modelname") tuples, and
when all the corresponding models have been imported and registered,
call the function with the model classes as its arguments. The function passed to this method must accept exactly n models as
arguments, where n=len(model_keys).
"""
# If this function depends on more than one model, we recursively turn
# it into a chain of functions that accept a single model argument and
# pass each in turn to lazy_model_operation.
model_key, more_models = model_keys[0], model_keys[1:]
if more_models:
supplied_fn = function def function(model):
next_function = partial(supplied_fn, model)
# Annotate the function with its field for retrieval in
# migrations.state.StateApps.
if getattr(supplied_fn, 'keywords', None):
next_function.field = supplied_fn.keywords.get('field')
self.lazy_model_operation(next_function, *more_models) # If the model is already loaded, pass it to the function immediately.
# Otherwise, delay execution until the class is prepared.
try:
model_class = self.get_registered_model(*model_key)
except LookupError:
self._pending_operations[model_key].append(function)
else:
function(model_class)

django之关联field 描述子是如何实现的的更多相关文章

  1. django之关联field 描述子

    """Accessors for related objects. When a field defines a relation between two models, ...

  2. django之relacted.py(探秘django的关联field)

    生成model类对象时,传入的每个field对象都会调用其contribute_to_class函数,生成对应的属性. def contribute_to_class(self, cls, name, ...

  3. django自关联,auth模块

    一.自关联 写蛮好的一篇博客:https://www.cnblogs.com/Kingfan1993/p/9936541.html 1.一对多关联 1.表内自关联是指表内数据相关联的对象和表是相同字段 ...

  4. BRIEF 特征描述子

    Binary Robust Independent Elementary Features www.cnblogs.com/ronny 1. BRIEF的基本原理 我们已经知道SIFT特征采用了128 ...

  5. 基于HOG-3D的时空描述子

    作者提出一种新的基于局部描述子的行为识别算法.

  6. Brief描述子

    一.Brief算法 1.基本原理 BRIEF是2010年的一篇名为<BRIEF:Binary Robust Independent Elementary Features>的文章中提出,B ...

  7. SIFT算法:特征描述子

    SIFT算法:DoG尺度空间生产  SIFT算法:KeyPoint找寻.定位与优化 SIFT算法:确定特征点方向  SIFT算法:特征描述子 目录: 1.确定描述子采样区域 2.生成描述子 2.1 旋 ...

  8. SIFT解析(三)生成特征描述子

    以上两篇文章中检测在DOG空间中稳定的特征点,lowe已经提到这些特征点是比Harris角点等特征还要稳定的特征.下一步骤我们要考虑的就是如何去很好地描述这些DOG特征点. 下面好好说说如何来描述这些 ...

  9. (二)ORB描述子提取源码思路与实现

    ORBSLAM2中ORB特征提取的特点 ORBSLAM2中通过对OpenCV中的ORB特征点提取类进行修改,对图像进行分块提取,而后划分节点,使得每个节点中保存的特征点性能是该节点所有特征点中最好的. ...

随机推荐

  1. JDK新特性之可变个数的形参的方法

    可变个数的形参的方法:1.格式:对于方法的形参来说:数据类型...形参名2.可变个数的形参的方法与同名的方法之间可以构成重载3.可变个数的形参在调用时,个数从0开始,到无穷都可以(在调用这个方法时,s ...

  2. HTC Vive设备拥有陀螺仪。

    //设置设备陀螺仪的开启/关闭状态,使用陀螺仪功能必须设置为 true Input.gyro.enabled = true; //获取设备重力加速度向量 Vector3 deviceGravity = ...

  3. consul & registrator & consul-template 使用

    consul & registrator & consul-template 使用 参考这里的文章: https://www.jianshu.com/p/a4c04a3eeb57 do ...

  4. Educational Codeforces Round 54 (Rated for Div. 2) D:Edge Deletion

    题目链接:http://codeforces.com/contest/1076/problem/D 题意:给一个n个点,m条边的无向图.要求保留最多k条边,使得其他点到1点的最短路剩余最多. 思路:当 ...

  5. Java(全局变量-静态变量-位运算符)

    全局变量是默认赋值的:而局部变量是没有默认赋值的(需要赋值才能使用)静态变量只能被静态方法使用 位运算右移,相当于做除法,2的n次幂00001000操作的位移数相当于是偏移量从右向左数n位,从第n+1 ...

  6. MySQL Hardware--CentOS 6查看CPU信息

    查看CPU信息 cpu_model=`cat /proc/cpuinfo |grep "model name"|awk -F ':' '{print $2}' |uniq` cpu ...

  7. HashTable & HashMap & ConcurrentHashMap 原理与区别

    一.三者的区别     HashTable HashMap ConcurrentHashMap 底层数据结构 数组+链表 数组+链表 数组+链表 key可为空 否 是 否 value可为空 否 是 否 ...

  8. IDEA下调试和运行Hadoop程序例子

    准备 配置好JDK和Hadoop环境, 在IDEA中建立maven项目,建立后的目录结构为: 修改pom..xml引入相关支持: <?xml version="1.0" en ...

  9. htm,css,javascript及其他的注释方式

    转自:http://www.cnblogs.com/dapeng111/archive/2012/12/23/2829774.html 一.HTML的注释方法<!-- html注释:START ...

  10. Druid密码加密

    pom里引用: <dependency> <groupId>com.alibaba</groupId> <artifactId>druid-spring ...