ANN: dbf (aka Python dBase)

[Ethan Furman]

The latest version, 0.95.001, is available on PyPI:

http://python.org/pypi/dbf


dbf v0.95.001
=============

dbf (also known as python dbase) is a module for reading/writing
dBase III, FP, VFP, and Clipper .dbf database files. It's
an ancient format that still finds lots of use (the most common
I'm aware of is retrieving legacy data so it can be stored in a
newer database system; other uses include GIS, stand-alone programs
such as Family History, Personal Finance, etc.).

Highlights
----------

Table -- represents a single .dbf/.dbt (or .fpt) file combination
and provides access to records; suports the sequence access and 'with'
protocols. Temporary tables can also live entirely in memory.

Record -- repesents a single record/row in the table, with field access
returning native or custom data types; supports the sequence, mapping,
attribute access (with the field names as the attributes), and 'with'
protocols. Updates to a record object are reflected on disk either
immediately (using gather() or write()), or at the end of a 'with'
statement.

Index -- nonpersistent index for a table.

Fields:
dBase III (Null not supported)

Character --> unicode
Date --> datetime.date or None
Logical --> bool or None
Memo --> unicode or None
Numeric --> int/float depending on field definition or None

Float --> same as numeric

Clipper (Null not supported)

Character --> unicode (character fields can be up to 65,519)

Foxpro (Null supported)

General --> str (treated as binary)
Picture --> str (treated as binary)

Visual Foxpro (Null supported)

Currency --> decimal.Decimal
douBle --> float
Integer --> int
dateTime --> datetime.datetime

If a field is uninitialized (Date, Logical, Numeric, Memo, General,
Picture) then None is returned for the value.

Custom data types:

Null --> used to support Null values

Char --> unicode type that auto-trims trailing whitespace, and
ignores trailing whitespace for comparisons

Date --> date object that allows for no date

DateTime --> datetime object that allows for no datetime

Time --> time object that allows for no time

Logical --> adds Unknown state to bool's: instead of True/False/None,
values are Truth, Falsth, and Unknown, with appropriate
tri-state logic with one caveat: as a matter of practicality
bool(Falsth) and bool(Unknown) are both False; if you want
bool(Unknown) to raise a TypeError instead, use Quantums.
__index__ of Unknown is 2, Truth is 1, and Falsth is 0.

Quantum --> similar to Logical, but implements boolean algebra (I think)


Whirlwind Tour
--------------

import datetime
import dbf

table = dbf.Table(':test:', 'name C(25); age N(3,0); birth D; qualified L')
table.open()

for datum in (
('Spanky', 7, dbf.Date.fromymd('20010315'), False),
('Spunky', 23, dbf.Date(1989, 07, 23), True),
('Sparky', 99, dbf.Date(), dbf.Unknown),
):
table.append(datum)

for record in table:
print record
print '--------'
print record[0:3]
print record['name':'birth']
print [record.name, record.age, record.birth]
print '--------'

custom = table.new(
filename='test_on_disk',
default_data_types=dict(C=dbf.Char, D=dbf.Date, L=dbf.Logical),
)

with custom: # automatically opened and closed
for record in table:
custom.append(record)
for record in custom:
dbf.write(record, name=record.name.upper())
print record
print '--------'
print record[0:3]
print record['name':'birth']
print [record.name, record.age, record.birth]
print '--------'

table.close()