Datasets

How datasets appear from a source, how to read their status, and how often they refresh.

A dataset is a named view over part of a source. Projects read datasets — never sources directly — which is what keeps access governed.

Where datasets come from

Connecting a source is enough. Chartizer introspects it and its contents appear on the Data sets tab automatically — tables and views for a relational database, indices for Elasticsearch, the parsed rows of a file. There is no separate step to opt a table in.

For query-defined connectors such as Neo4j, the dataset is the query you write, and its returned columns are the fields.

Each dataset has:

a name, unique within the team,
a type — Table, View, Index, Query, or File,
a schema — field names and types, captured on the first sync,
a refresh policy.

Status reference

Status	Colour	Meaning
`Ready`	green	Last sync succeeded; schema matches
`Syncing`	blue	Refresh running now
`Auth error`	red	Parent source cannot authenticate
`Schema drift`	amber	The source’s shape changed since the last sync
`Stale`	grey	Source paused; showing last synced data

Status belongs to the dataset, cause often belongs to the source

Auth error on four datasets is usually one broken source. Fix it once at the source.

Refresh policies

Policy	Behaviour	Good for
Live query	Runs against the source at chart render	Small tables, always-fresh numbers
Scheduled	Syncs on an interval (15 min → daily)	Large tables, rate-limited APIs
On upload	Refreshes when a new file replaces the old	CSV / Excel
Manual	Only when someone clicks Sync now	Frozen reference data

Scheduled refresh, every 15 minutes
{
  "refresh": {
    "policy": "scheduled",
    "interval": "15m",
    "timezone": "Europe/Riga",
    "on_failure": "retry_then_alert"
  }
}

Live queries inherit source latency

A live-query dataset on a slow database makes every chart that reads it slow. If a chart takes more than a few seconds to render, switch that dataset to scheduled.

Schema drift

When a sync finds a schema that differs from the one it recorded, the dataset moves to Schema drift instead of failing.

New columns are available immediately but excluded from existing charts; accept the change to add them to the schema. A whole new table in the source appears as a new dataset the same way — nothing to opt in to.

Charts referencing the removed field show a broken-field marker. Either restore the column at the source or edit the chart.

The most disruptive case — integer becoming text breaks aggregation. Chartizer holds the last good sync until you accept the change.

Sync history

Every dataset keeps its recent syncs: start time, duration, row count, and outcome. A sudden change in row count is usually the first sign of an upstream problem.

Text

Today 14:02   ✓ 12,480 rows   4.1s
Today 13:47   ✓ 12,477 rows   3.9s
Today 13:32   ✗ auth error    0.2s
Today 13:17   ✓ 12,455 rows   4.0s

Delete a dataset

Deleting removes the dataset and its mappings. The Projects column tells you how many projects read it first — if that number is not zero, check with their owners.