Modelling icons

The 50 modelling icons in Data Engineering, described

Audit tableaudit-table

An audit table — who changed what, kept beside the table it watches.

Bitemporalbitemporal

Bitemporal — when a thing was true, and when the table was told about it.

Data branchbranch-data

A data branch — a copy of the warehouse to change without touching what is live.

Change log tablechange-log-table

A change log table — one row per change, kept so the state can be rebuilt.

Clustering keyclustering-key

A clustering key — the column the rows are physically sorted by, so a scan stays small.

Columnar storecolumnar-store

A columnar store — the values of one column kept together, so a scan reads only what it needs.

Conformed dimensionconformed-dimension

A conformed dimension — one dimension meaning the same thing in every mart that uses it.

Copy intocopy-into

Copy into — files picked up from a stage and written into the table.

Data martdata-mart

A data mart — one team's slice of the warehouse, shaped for the questions they ask.

Data modeldata-model

A data model — the entities and the lines between them, before any of it is a table.

Data vaultdata-vault

A data vault — hubs, links and satellites, so history is added and never edited.

Model filedbt-model

A model file — the select that becomes a table, and the tests that come with it.

Denormalisedenormalise

Denormalise — the joins done once and stored, so the reader never has to do them.

Dimension tabledimension-table

A dimension table — the labels a fact is described by, and nothing that happened.

Effective dateeffective-date

An effective date — the day a row starts being the true one.

Entity relationentity-relation

An entity relation — one row on this side, many on that, written on the line.

External tableexternal-table

An external table — a table the warehouse reads but does not own.

Fact tablefact-table

A fact table — the measurements, one row per thing that happened.

Federated queryfederated-query

A federated query — one statement answered from more than one system at once.

File formatfile-format

A file format — how the rows are written on disk, which decides what a read costs.

Flattenflatten-nested

Flatten — a nested shape opened out into ordinary columns.

Full refreshfull-refresh

A full refresh — the table dropped and rebuilt from the source every time.

Graingrain-table

The grain — what one row of the table means, decided before anything is loaded.

Hard deletehard-delete

A hard delete — the row actually gone, and nothing left to read.

Idempotent loadidempotent-load

An idempotent load — running it twice leaves the table exactly as running it once.

Incremental modelincremental-model

An incremental model — only the rows that arrived since last time are built.

Junk dimensionjunk-dimension

A junk dimension — the leftover flags gathered into one table so the fact stays clean.

Late-arriving datalate-arriving-data

Late-arriving data — a row for a window that was already closed and counted.

Column lineagelineage-column

Column lineage — which column fed which, through every model between.

Table lineagelineage-table

Table lineage — which tables this one was built from, and which are built from it.

Manifestmanifest-file

A manifest — the file that lists which files are actually part of the table right now.

Merge intomerge-into

Merge into — matched rows updated, new ones inserted, in one statement.

Migration downmigration-down

A migration down — the change undone, the schema put back as it was.

Migration upmigration-up

A migration up — the change applied, the schema moved to the next version.

Nested fieldnested-field

A nested field — a column that holds a shape rather than a value.

Partition pruningpartition-prune

Partition pruning — the partitions the query cannot need are never opened.

Row storerow-store

A row store — a whole row kept together, which is what a transaction wants.

Schema registryschema-registry

A schema registry — every version of every schema, kept where producers can check it.

Schema versionschema-version

A schema version — which shape of the table this is, and when it became that.

Slowly changing dimensionslowly-changing-dimension

A slowly changing dimension — the old row kept when a label changes, so history still reads right.

Snowflake schemasnowflake-schema

A snowflake schema — the dimensions of a star, themselves broken into more.

Soft deletesoft-delete

A soft delete — the row marked gone and kept, so nothing downstream loses its history.

Staging tablestaging-table

A staging table — where the raw load lands before anything is done to it.

Star schemastar-schema

A star schema — one fact table in the middle, its dimensions around it.

Struct columnstruct-column

A struct column — named fields kept inside one column of the row.

Surrogate mapsurrogate-map

A surrogate map — the table that says which business key became which generated id.

Time traveltime-travel-query

Time travel — the table read as it stood at a moment that has passed.

Unloadunload-data

Unload — the table written back out to files a system outside can read.

Watermark columnwatermark-column

A watermark column — the timestamp a load reads to know where it stopped.

Wide tablewide-table

A wide table — everything flattened into one row so nothing has to be joined.

In code, each is one import — import { AuditTable } from "@iconmind/react/icons/audit-table" — and the same name in Vue, Svelte, Solid, Preact, React Native, Astro, Blade and Flutter.

Narrow it by tag