Multiple-file datasource

Combine files with matching columns into one logical datasource. Complete local setup, then download sales-jan.csv, sales-feb.csv, multiple-files.preql, and glob-files.preql into one folder.

January contains sales 1 and 2 (store: 120; web: 80). February contains sales 3 and 4 (store: 50; web: 30). Sale IDs are unique across both files.

Explicit paths

The list makes the set of input files explicit. All paths must have the same file extension.

key sale_id int;
property sale_id.channel string;
property sale_id.amount float;

datasource sales (
    sale_id: sale_id,
    channel: channel,
    amount: amount
)
grain (sale_id)
file [`./sales-jan.csv`, `./sales-feb.csv`];

select channel, sum(amount) as revenue
order by channel asc;

trilogy run multiple-files.preql duckdb

Glob pattern

Use a glob when every matching file belongs to the dataset. Keep unrelated CSVs out of the matched set.

key sale_id int;
property sale_id.channel string;
property sale_id.amount float;

datasource sales (
    sale_id: sale_id,
    channel: channel,
    amount: amount
)
grain (sale_id)
file `./sales-*.csv`;

select channel, sum(amount) as revenue
order by channel asc;

trilogy run glob-files.preql duckdb

Both recipes return:

channel	revenue
store	170
web	110

These bindings combine rows; they do not deduplicate overlapping files or enforce key uniqueness. For separate datasource declarations with scoped completeness, see partial sources.