duckdb community extension
dta
read and write Stata
.dta files, natively in DuckDB.
dta is a DuckDB extension that reads and writes Stata .dta files — formats 117 through 121, corresponding to Stata 13–18 — with full type mapping, value labels, and strL support.
install in 10 seconds
dta is distributed through the DuckDB community extension repository. From any DuckDB session:
SQL// in any duckdb session
INSTALL dta FROM community;
LOAD dta;usage
The extension adds the table function read_dta, registers itself as the default reader for .dta files, and enables writing tables to .dta with COPY.
SQL
-- Read a .dta file
SELECT * FROM read_dta('auto.dta');
-- Or let DuckDB infer the format from the file name
SELECT * FROM 'auto.dta';
-- Read with value labels converted to DuckDB ENUMs
SELECT * FROM read_dta('auto.dta', value_labels=true);
-- Write a table to .dta
COPY my_table TO 'output.dta' (FORMAT dta);
-- The default writer also infers .dta from the file name
COPY my_table TO 'output.dta';why dta
Native, both directions
Read and write
.dta inside DuckDB — no Stata installation, no Python round-trip, no CSV export in between.Dates and datetimes
Stata
%td and %tc formats convert to DuckDB DATE and TIMESTAMP — and back again on write.Value labels
Pass
value_labels=true to get labeled columns as DuckDB ENUMs. On write, ENUM columns become value labels again.Long strings
Fixed-length strings and
strL both read as VARCHAR; long strings write back as strL entries.type mapping
reading — .dta to DuckDB
| Stata type | DuckDB type |
|---|---|
byte | TINYINT |
int | SMALLINT |
long | INTEGER |
float | FLOAT |
double | DOUBLE |
double (%td) | DATE |
double (%tc) | TIMESTAMP |
strN, strL | VARCHAR |
writing — DuckDB to .dta
| DuckDB type | Stata type |
|---|---|
BOOLEAN, TINYINT | byte |
SMALLINT | int |
INTEGER | long |
BIGINT, HUGEINT | double |
FLOAT | float |
DOUBLE, DECIMAL | double |
DATE | double (%td) |
TIMESTAMP | double (%tc) |
VARCHAR | strL |
ENUM | byte/int/long + value labels |
output format
The writer produces format 119 (Stata 15), supporting up to 2 billion variables. When reading with value_labels=true, columns with value labels are returned as ENUM.