feat(ava/lite-insight): 新增多维度下钻波动归因方法 #658
Conversation
GuangMingYouBei
force-pushed
the
v3-fluctuation-youbei
branch
from
a9a9325
to
cdc2745
Compare
GuangMingYouBei
force-pushed
the
v3-fluctuation-youbei
branch
from
cdc2745
to
81e2e7d
Compare
| @@ -124,6 +124,7 @@ export type { | |||
| LowVarianceInfo, | |||
| CorrelationInfo, | |||
| InsightsResult, | |||
| DimensionDrillDownProps, | |||
There was a problem hiding this comment.
改为了 options 的格式。
There was a problem hiding this comment.
改为了 options 的格式。
这是改了之后的还是没改的😂
| @@ -93,7 +93,7 @@ export type { | |||
| } from './data'; | |||
|
|
|||
| /* insight */ | |||
| export { getInsights, generateInsightVisualizationSpec } from './insight'; | |||
| export { getInsights, generateInsightVisualizationSpec, dimensionDrillDownAttribution } from './insight'; | |||
There was a problem hiding this comment.
dimensionDrillDownAttribution 需要单独导出么，目前其他算法都是通过 getInsights 指定 insightType 导出的。单独导出 dimensionDrillDownAttribution 不太符合统一的调用形式
There was a problem hiding this comment.
暂时去掉了单独导出，但是感觉这种算子是有可能单独调用的，以后再考虑怎么导出。
| @@ -1,3 +1,5 @@ | |||
| export { getInsights } from './pipeline'; | |||
| export { generateInsightVisualizationSpec } from './pipeline/visualize'; | |||
| export { dimensionDrillDownAttribution } from './insights/extractors/causalInference/dimensionDrillDown'; | |||
| export type { DrillDownProps as DimensionDrillDownProps } from './insights/extractors/causalInference/types'; | |||
There was a problem hiding this comment.
同上，确认下 dimensionDrillDownAttribution 是否可以不单独导出
另外 types 如果需要导出，写到 types 文件下，下面使用了 export * from './types' 导出 types
| const tempResult = dimensionDrillDownAttribution(props); | ||
| const treeResult = tempResult.resultInTree; | ||
| expect(treeResult?.City.huangshishi.info.currValue).toBe(13722.76); | ||
| }); |
There was a problem hiding this comment.
除了 currValue, 还有 diff, 贡献度等结果需要在测试用例中体现？
|
|
||
| /** DataConfig specifies the input data with its focused dimensions and target measure to be analysed. */ | ||
| export type DataConfig = { | ||
| sourceData: Datum[]; |
There was a problem hiding this comment.
这些信息部分在 InsightOptions 里已经定义了，建议走 getInsights 整体的流程，不重复指标、维度这些参数
| }; | ||
|
|
||
| /** Dimension drill down Result type */ | ||
| export type DimensionDrillDownResult = { |
There was a problem hiding this comment.
类型命名和结果的结构整体改一下，和 insight 中其他算子类似，结果使用 BasePatternInfo 的命名和数据结构，可参考 TrendInfo
算法入参如 FluctInfo 可以统一叫 xxOptions
| targetMeasure: string; | ||
| } | ||
|
|
||
| export interface MeasureDecomposeProps extends CausalInferenceProps { |
There was a problem hiding this comment.
这个文件里有的地方用的 interface，有的用 type，不太统一
如果不是必须要用 interface，可以都用 type
| @@ -0,0 +1,128 @@ | |||
| import { cloneDeep as _cloneDeep } from 'lodash'; | |||
There was a problem hiding this comment.
这个文件里的工具函数可以看下有没有可以放到 insight/utils 目录下面的（比较通用的可放）
| measure: string, | ||
| fluctuationDim: string, | ||
| location: DataLocation | ||
| ) => { |
There was a problem hiding this comment.
- 这个函数参数有点多，建议用 object 管理，否则调用时不太方便一一对应上
- 需要加上注释说明函数和每个参数含义，比如 item, dictFlatten, deque, location 都不是很好理解
- 命名修改？ dictFlatten --> flattenDict, fluctuationDim --> fluctuationDimension
| globalDiff.diff = globalDiff.currValue - globalDiff.baseValue; | ||
|
|
||
| return { resultInTree: resultTree, globalDiff, resultInList: Object.values(DictFlatten) }; | ||
| }; |
There was a problem hiding this comment.
建议参考其他算子包一个 extractor function，这样外层可以用统一的方法调用各种类型的 insight extractor
| @@ -0,0 +1,75 @@ | |||
| import { dimensionDrillDownAttribution } from '../../../../../src/insight/insights/extractors/causalInference/dimensionDrillDown'; | |||
There was a problem hiding this comment.
commit message : feat(ava/lite-insight) --> feat(ava/insight)，且用英文写 commit message。
补充下 pr 描述
No description provided.