Advanced Data Dictionary

未关闭
  • 哪里不对？
#18 13 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
8/100
Issue 类型
功能
描述清晰度
需要澄清
活跃度
冷清
技术栈
postgresql, python
领域
data, database

调研方向

This is an open design question, not a code task: the issue lists six storage options for qsv statistics and frequency tables and does not choose one. No files or tests are named. Before any work, a maintainer (the issue asks @wardi) needs to pick an option. Done would mean a recorded decision and a scoped first step, not an implementation.

由索引模型根据 Issue 内容生成。

描述

enhancement

Currently, CKAN's Data Dictionary is limited to data type, label and description.

With qsv stats we collect descriptive statistics when we infer each column's data type during the Analysis phase of a DP+ job.

Currently - sum, min/max, min/max length, mean, stddev, variance, quartiles, median, modes, cardinality & nullcount.

When cardinality = rowcount and nullcount = 0, we can infer that a column can be a primary key and be a unique index, and annotate its data dictionary accordingly (and going further, create a unique index on it after the Copy phase).

When nullcount = rowcount, we can infer that a column is empty, and note it in the data dictionary as well.

And with qsv frequency - we can also compile frequency tables for the top N values of a column, and if the cardinality of a column is below a given N threshold, we can even infer the domain of a column as enumerated values.

Since we paid for compiling the statistics when we inferred the column data types already, we can store these statistics in the data dictionary as well as "schemata" (a term I coined for schema metadata) for "free" ( or nearly free, as running qsv frequency is not currently done by DP+, but even against a large file like the 1m row, 500mb NYC 311 benchmark data, it only takes 2.8 seconds).

We have several options:

  1. add additional properties to the existing Data Dictionary JSON which is stored as a table comment
  2. keep the existing Data Dictionary JSON as is, and store RESOURCE_NAME-stats.csv and RESOURCE_NAME-freq.csv as "system resources", that can be downloaded and queried with the CKAN API.
  3. extend the Data Dictionary with additional properties, and also store the -stats and -freq CSVs as system resources.
  4. alternatively, instead of using the original Data Dictionary JSON, we can instead insert the jsonschema file produced by the qsv schema command. The added benefit of doing so is that we use the jsonschema file with qsv validate to check if an external file conforms to the schema. And since qsv validate accepts a jsonschema URL, you can even validate an external file against the CKAN hosted jsonschema.
  5. Do 4, and add the "system resources" like 3.
  6. Store all these schemata data in a "schemata catalog" in the datastore database or a dedicated schemata database as native PostgreSQL objects. Perhaps, by using a resource's ID and adding a special prefix and/or suffix to it (e.g. RESOURCEID_datadict, RESOURCEID_stats, RESOURCEID_freq). Doing so has the added benefit of being able to query all the data dictionaries - e.g. columns with the same name, infer related resources, suggest joins, etc. and other "Linked Data" like queries and inferences in a performant manner.

@wardi , since you originally implemented the Data Dictionary, would be keen to get your opinion.

主要语言
Python
星标
47
派生
33
平均合并
2 小时 28 分钟
30 天内合并 PR
10

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

dathere/datapusher-plus 的其他 Issue

查看 dathere/datapusher-plus 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。