When the need arises to document database schemas, different DBMSs provide their own tools for similar tasks. And most of them support DESC table_name
, including ClickHouse. However, the result of this command is not as expressive as we would like.
DESCRIBE TABLE data_lr name type default_type default_expression comment codec_expression ttl_expression Path String ZSTD(3) Value Float64 Gorilla, LZ4 Time UInt32 DoubleDelta, LZ4 Date Date DoubleDelta, LZ4 Timestamp UInt32 DoubleDelta, LZ4
At the same time, the system tables tables
and columns
contain comprehensive information, combining that you can get here is a pretty result:
Inspiration
Before starting to carefully saw my own solution, devoid of fatal flaws, I looked for available options. And this work inspired me to start my own project. Some blanks were gratefully borrowed from there, with attribution.
What is PlantUML
A few words about what this product is. For the best understanding, I advise you to go to the official page . In a nutshell, this is a program that converts the textual description of diagrams, for example:
@startuml Bob -> Alice : hello @enduml
in images
, , , , . Atlassian Confluence, wiki. pandoc ( ), LaTeX, .
: system.tables , , , , . , system.columns, . , , . .
, , . , . , ReplacingMergeTree Version. . Replicated*MergeTree. MaterializedView . , MV.
. , ± . -. .
. , Distributed , Buffer *MergeTree
, . . .
@startuml ' This diagram is generated with https://github.com/Felixoid/clickhouse-plantuml !define Table(x) class x << (T,mistyrose) >> !define View(x) class x << (V,lightblue) >> !define MaterializedView(x) class x << (m,orange) >> !define Distributed(x) class x << (D,violet) >> hide empty methods hide stereotypes skinparam classarrowcolor gray Distributed(graphite.data) { ENGINE=**Distributed** ..engine config.. cluster: graphite_data database: graphite table: data_lr sharding_key: cityHash64(Path) ==columns== Path: String Value: Float64 Time: UInt32 Date: Date Timestamp: UInt32 } Table(graphite.data_lr) { ENGINE=**ReplicatedGraphiteMergeTree** ..engine config.. rollup_config: graphite_rollup ..replication.. zoo_path: /clickhouse/tables/graphite.data_lr/{shard} replica: {replica} ==columns== Path: String <size:15><&signal></size> Value: Float64 Time: UInt32 <size:15><&signal></size> Date: Date <size:15><&list-rich></size> Timestamp: UInt32 ..<size:15><&list-rich></size>partition key.. toYYYYMMDD(toStartOfInterval(Date, toIntervalDay(3))) ..<size:15><&signal></size>sorting key.. Path, Time } Table(graphite.index) { ENGINE=**ReplicatedReplacingMergeTree** ..engine config.. version: Version ..replication.. zoo_path: /clickhouse/tables/graphite.index/1 replica: {replica} ==columns== Date: Date <size:15><&list-rich></size> <size:15><&signal></size> Level: UInt32 <size:15><&signal></size> Path: String <size:15><&signal></size> Version: UInt32 ..<size:15><&list-rich></size>partition key.. toYYYYMM(Date) ..<size:15><&signal></size>sorting key.. Level, Path, Date } Table(graphite.tagged) { ENGINE=**ReplicatedReplacingMergeTree** ..engine config.. version: Version ..replication.. zoo_path: /clickhouse/tables/graphite.tagged/1 replica: {replica} ==columns== Date: Date <size:15><&list-rich></size> <size:15><&signal></size> Tag1: String <size:15><&signal></size> Path: String <size:15><&signal></size> Tags: Array(String) Version: UInt32 ..<size:15><&list-rich></size>partition key.. toYYYYMM(Date) ..<size:15><&signal></size>sorting key.. Tag1, Path, Date } graphite.data_lr -|> graphite.data @enduml
, — ClickHouse. ClickHouse. , - issue. , .
, - .
, pypi:
pip install clickhouse-plantuml clickhouse-plantuml -h