Packages

Read DBASE files in Elixir. At the moment it only supports read.

Current section

Files

Jump to
dbf_ex README.md
Raw

README.md

# DBF
Read DBASE files in Elixir.
At the moment it only supports read.
## Usage
For callback-scoped reads, use `DBF.with_open/2,3`. It closes the DBF and any
memo resource after the callback returns or raises:
```elixir
records =
DBF.with_open("test/dbf_files/bayarea_zipcodes.dbf", fn db ->
Enum.to_list(db)
end)
```
An open database implements `Enumerable`. Each element is a `{status, values}`
tuple whose status is `:record` or `:deleted_record`.
Use `DBF.get/2` for a specific zero-based record index:
```elixir
DBF.with_open("test/dbf_files/bayarea_zipcodes.dbf", fn db ->
case DBF.get(db, 2) do
{:record, row} -> IO.inspect(row)
{:deleted_record, row} -> IO.inspect(row)
{:error, error} -> IO.warn(Exception.message(error))
end
end)
```
For streaming, suspended enumeration, or longer-lived access, use
`DBF.open/1,2` and pair every successful open with `DBF.close/1`.
### Exact numeric values
Numeric fields remain floats by default for compatibility. Opt into exact values
to receive integers for scale-zero fields and `Decimal` values for positive
scales:
```elixir
DBF.with_open("table.dbf", [numeric: :exact], fn db ->
DBF.get(db, 0)
end)
```
Malformed and blank numeric fields remain `nil` under this value policy.
### Text encoding
Known language drivers for Windows-1251 and Windows-1252 are decoded to UTF-8.
Missing or unknown drivers preserve raw bytes by default instead of guessing:
```elixir
DBF.with_open(
"table.dbf",
[encoding: :windows_1251, encoding_errors: :strict],
fn db -> Enum.to_list(db) end
)
```
`encoding` accepts `:auto`, `:raw`, `:windows_1251`, or `:windows_1252`.
`encoding_errors` accepts `:strict`, `:replace`, or `:raw`. The defaults are
`:auto` and `:raw`. The selected policy applies consistently to field names,
character values, and textual DBT memos; binary values are not decoded as text.
## Format compatibility
Support is evidence-based and applies only to the capabilities exercised by the
checked-in fixtures. A recognized version byte alone does not imply support.
- **Verified** — representative headers, records, and applicable memo values are
covered by local expected values.
- **Partial** — some real files work, but known format features or value semantics
remain incomplete.
- **Planned** — fixtures and primary references are recorded, but the format is
not accepted yet.
- **Not planned** — outside the scope of the read-only table reader.
| Format/profile | Version bytes | Level | Notes |
| ----------------------------------- | ------------------------------ | ----------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| FoxBase | `0x02` | Verified | All fixture records and `C`/unscaled `N` values are checked, including blanks, exact numerics, and deleted-record behavior. |
| dBASE III without memo | `0x03` | Verified | All zipcode oracle rows, the full schema, legacy value states, exact numerics, Windows-1252 policies, and ambiguous-schema rejection are covered. |
| dBASE III with DBT memo | `0x83` | Verified | Representative complete records, schema, exact numerics, logical states, multi-block memos, pointers, encoding overrides, and companion validation are covered. |
| dBASE IV with DBT memo | `0x8B` | Verified | Representative schema and values, text policies, declared block sizing, multi-block memos, and companion validation are covered. |
| FoxPro and Visual FoxPro tables/FPT | `0x30`, `0x31`, `0x32`, `0xF5` | Planned | Fixtures cover FPT, autoincrement, variable-width fields, null flags, and CP1251 text. |
| dBASE Level 7-style tables | `0x8C` fixture | Planned | Extended header/descriptor and memo support are not implemented. |
| DBF writing || Not planned | Read-only scope. |
| NDX/MDX/CDX/DCX index reading || Not planned | Tracked separately from table reading. |
See `test/support/fixture_manifest.ex` for per-fixture provenance, encoding,
redistribution status, expected-value source, and normative references.