根据数据创建规格 [英] create spec from data

问题描述

我正在尝试仅根据数据创建规范。我的数据结构非常复杂-所有嵌套的地图。

I am trying to create spec just from data. I have very complex data structure - all nested map.

{:contexts
 ({:importer.datamodel/global-id "01b4e69f86e5dd1d816e91da27edc08e",
   :importer.datamodel/type "province",
   :name "a1",
   :importer.datamodel/part-of "8cda1baed04b668a167d4ca28e3cef36"}
  {:importer.datamodel/global-id "8cda1baed04b668a167d4ca28e3cef36",
   :importer.datamodel/type "country",
   :name "AAA"}
  {:importer.datamodel/global-id "c78e5478e19f2d7c1b02088e53e8d8a4",
   :importer.datamodel/type "location",
   :importer.datamodel/center ["36." "2."],
   :importer.datamodel/part-of "01b4e69f86e5dd1d816e91da27edc08e"}
  {:importer.datamodel/global-id "88844f94f79c75acfcb957bb41386149",
   :importer.datamodel/type "organisation",
   :name "C"}
  {:importer.datamodel/global-id "102e96468e5d13058ab85c734aa4a949",
   :importer.datamodel/type "organisation",
   :name "A"}),
 :datasources
 ({:importer.datamodel/global-id "Source;ACLED",
   :name "ACLED",
   :url "https://www.acleddata.com"}),
 :iois
 ({:importer.datamodel/global-id "item-set;ACLED",
   :importer.datamodel/type "event",
   :datasource "Source;ACLED",
   :features
   ({:importer.datamodel/global-id
     "c74257292f584502f9be02c98829d9fda532a492e7dd41e06c31bbccc76a7ba0",
     :date "1997-01-04",
     :fulltext
     {:importer.datamodel/global-id "df5c7d6d075df3a7719ebdd39c6d4c7f",
      :text "bla"},
     :location-meanings
     ({:importer.datamodel/global-id
       "e5611219971164a15f06e07228fb7b51",
       :location "8cda1baed04b668a167d4ca28e3cef36",
       :contexts (),
       :importer.datamodel/type "position"}
      {:importer.datamodel/global-id
       "af36461d27ec1d8d28fd7f4a70ab7ce2",
       :location "c78e5478e19f2d7c1b02088e53e8d8a4",
       :contexts (),
       :importer.datamodel/type "position"}),
     :interaction-name "Violence",
     :importer.datamodel/type "description",
     :has-contexts
     ({:context "102e96468e5d13058ab85c734aa4a949",
       :context-association-type "actor",
       :context-association-name "actor-1",
       :priority "none"}
      {:context "88844f94f79c75acfcb957bb41386149",
       :context-association-type "actor",
       :context-association-name "actor-2",
       :priority "none"}),
     :facts
     ({:importer.datamodel/global-id
       "c46802ce6dcf33ca02ce113ffd9a855e",
       :importer.datamodel/type "integer",
       :name "fatalities",
       :value "16"}),
     :attributes
     ({:name "description",
       :importer.datamodel/type "string",
       :value "Violence"})}),
   :attributes (),
   :ioi-slice "per-item"})}

什么样的工具可以创建这种结构的规范？
我正在尝试使用此工具： https://github.com/stathissideris/spec-提供商

What tool can create the spec for such a structure? I am trying to use this tool: https://github.com/stathissideris/spec-provider

，但是它给了我：

(spec/def :importer.datamodel/data
  (clojure.spec.alpha/coll-of
   (clojure.spec.alpha/or
    :collection
    (clojure.spec.alpha/coll-of
     (clojure.spec.alpha/keys
      :req
      [:importer.datamodel/global-id]
      :opt
      [:importer.datamodel/center
       :importer.datamodel/part-of
       :importer.datamodel/type]
      :opt-un
      [:importer.datamodel/attributes
       :importer.datamodel/datasource
       :importer.datamodel/features
       :importer.datamodel/ioi-slice
       :importer.datamodel/name
       :importer.datamodel/url]))
    :simple
    clojure.core/keyword?)))

这不是完整的解决方案离子...
我用（sp / pprint-specs（sp / infer-specs data：importer.datamodel / data）'data'） ..
什么工具可以为这种结构创建规格？

which is not complete solution... I use (sp/pprint-specs (sp/infer-specs data :importer.datamodel/data) 'data 's)... What tool can create the spec for such a structure?

推荐答案

我是尝试使用此工具： https://github.com/stathissideris/spec-provider

规范提供者无法提供所需的结果，因为您的数据是复杂的嵌套/递归结构。其中一些地图最好与多规格规范，但会获得规范提供商不这样做；文档中的一项警告指出 没有尝试推断多规格。

spec-provider isn't giving you the desired result because your data is a complex nested/recursive structure. Some of those maps would be best spec'd with multi-specs, but spec-provider won't do that; one of the caveats in its docs says There is no attempt to infer multi-spec.

唯一的方法要正确指定其中一些地图正在使用多规范，其规范将取决于其：importer.datamodel / type 值。

The only way to properly spec some of these maps is using multi-specs their spec will depend on their :importer.datamodel/type value.

首先，让我们看一下顶级键（假设地图位于名为 data 的绑定中）：

First, let's look at the top-level keys (assuming the map is in a binding named data):

(keys data) => (:contexts :datasources :iois)

创建 s /密钥最外层地图的规格：

Create a s/keys spec for the outermost map:

(s/def ::my-map
  (s/keys :req-un [::contexts ::datasources ::iois]))

这些键是不合格的，但是我们必须使用带有：req-un 的合格关键字来指定它们。我们可以使用REPL通过遍历嵌套结构并收集数据来查看嵌套地图的形状及其与：importer.datamodel / type 的关系：

These keys are unqualified, but we must use qualified keywords w/:req-un to spec them. We can use the REPL to look at the shapes of nested maps and their relationships to :importer.datamodel/type, by walking the nested structure and collecting data:

(let [keysets (atom #{})]
  (clojure.walk/postwalk
    (fn [v]
      (when (map? v)
        (swap! keysets conj [(:importer.datamodel/type v) (keys v)]))
      v)
    data)
  @keysets)
=>
#{...
  ["organisation" (:importer.datamodel/global-id :importer.datamodel/type :name)]
  [nil (:context :context-association-type :context-association-name :priority)]
  ["description"
   (:importer.datamodel/global-id :date :fulltext :location-meanings
    :interaction-name :importer.datamodel/type :has-contexts :facts :attributes)]
  ["event" (:importer.datamodel/global-id :importer.datamodel/type :datasource :features :attributes :ioi-slice)]
 ...}

（即将发布的规范alpha应该可以更轻松地从程序中定义规范

(An upcoming spec alpha should make it easier to define specs programmatically from this data.)

我们可以看到有些地图形状没有一个：importer.datamodel / type ，但是我们可以为那些编写多规范。首先定义用于分派类型键的多重方法：

We can see there are some map shapes that don't have a :importer.datamodel/type, but we can write multi-specs for the ones that do. First define a multimethod for dispatching on the type key:

(defmulti type-spec :importer.datamodel/type)

然后为每个 defmethod >：importer.datamodel / type 值。以下是一些示例：

Then write a defmethod for each :importer.datamodel/type value. Here are a few examples:

(defmethod type-spec :default [_] (s/keys))
(defmethod type-spec "organisation" [_]
  (s/keys :req [:importer.datamodel/global-id]
          :req-un [::name]))
(defmethod type-spec "description" [_]
  (s/keys :req [:importer.datamodel/global-id]
          :req-un [::date ::fulltext ::location-meanings ::interaction-name
                   ::has-contexts ::facts ::attributes]))
(defmethod type-spec "event" [_]
  (s/keys :req-un [::features]))

然后定义 s / multi-规范：

(s/def ::datamodel
  (s/multi-spec type-spec :importer.datamodel/type))

现在我们符合的任何地图 :: datamodel 将根据其：importer.datamodel / type 值解析规范。我们可以将该规范分配给规范将用于符合地图要求的关键字，例如最外面的键之一：

Now any map we conform to ::datamodel will resolve a spec based on its :importer.datamodel/type value. We can assign that spec to keywords that spec will use to conform the maps, e.g. one of the outermost keys:

(s/def ::contexts (s/coll-of ::datamodel))

现在，如果您从我们在下指定的一张地图中删除所需的密钥：contexts ，规范可以告诉您出了什么问题。例如，从组织 映射中删除：name 键：

Now if you remove a required key from one of the maps we spec'd under :contexts, spec can tell you what's wrong. For example, removing the :name key from an "organisation" map:

(s/explain ::my-map data)
In: [:contexts 3]
val: #:importer.datamodel{:global-id "88844f94f79c75acfcb957bb41386149",
                          :type "organisation"}
fails spec: :playground.so/datamodel
at: [:contexts "organisation"]
predicate: (contains? % :name)

其他规格

对于没有：importer.datamodel / type 的地图，您应该能够定义关键规格。例如，嵌套的：has-contexts 键具有一个没有：importer.datamodel / type 的地图集合，但是，如果我们可以假设它们都相似，我们可以编写以下规范：

Other specs

For the maps that don't have a :importer.datamodel/type you should be able to define a key spec. For example, the nested :has-contexts key has a collection of maps without a :importer.datamodel/type, but if we can assume they'll all be similar we can write this spec:

(s/def ::has-contexts
  (s/coll-of (s/keys :req-un [::context ::context-association-type
                              ::context-association-name ::priority])))

：具有上下文在我们的地图中上面已经包含了多规范，只需将规范注册到该密钥即可使规范符合其值。包含此规范的最外键是：iois ，因此我们也可以指定该键：

:has-contexts is in a map we've already covered with a multi-spec above, and simply registering a spec to this key will make spec conform its values. The outermost key that contains this spec is :iois so we can spec that key too:

(s/def ::iois (s/coll-of ::datamodel))

现在，将输入符合 :: my-map 规范将自动覆盖更多数据。

Now, conforming an input to ::my-map spec will automatically cover more data.

什么工具可以创建这种结构的规范？

What tool can create the spec for such a structure?

如您所见，为此编写了完整的规范结构是不平凡的，但可能的。我不知道任何现有工具可以自动推断出此结构的完整正确规格。不得不直觉：importer.datamodel / type 是一个可以用于分派给不同的 s / keys 规范-它仍然会做出可能无效的假设。我认为在这种情况下，工具-辅助规范生成更加现实和实用。

As you can see, writing a full spec for this structure is non-trivial but possible. I don't know of any existing tool that could automatically infer a complete, "correct" spec for this structure. It would've had to intuit that :importer.datamodel/type is a key that could be used to dispatch to different s/keys specs — and it would still be making a potentially invalid assumption. I think tool-assisted spec generation is more realistic and practical in this case.

这篇关于根据数据创建规格的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！