DBF to CSV

Why turn a DBF into CSV

DBF tables tend to outlive the programs that made them. A stock system written in FoxPro, an accounting package from the DOS years, the attribute table of a shapefile: the data is still useful, but the program that opened it may be long gone. CSV is the plainest way to carry that data forward. Databases, scripts and spreadsheets all read it, and because it's just text, it will still open years from now.

Typical jobs: moving customer or product records into a new system that imports CSV, loading a map layer's attributes into a database, or simply looking inside a table that nothing on your computer opens anymore.

How to convert a DBF file to CSV

1. Above the drop area, check the two settings: Delimiter (Comma to start with) and Text encoding (From the file). A setting applies to the files you add after choosing it.
2. Drop your .dbf onto the page, or press Choose file. If the table has memo fields, select its .dbt or .fpt file in the same go; it's matched to the table by name.
3. Each file gets its own row. When it's done, open What changed under the file name to see which encoding was used and anything that was left out.
4. Press Download. With several files, Download all (.zip) saves them together. The free plan converts up to 5 files at a time.

Almost everything in a DBF comes through untouched. What we leave out, we leave out on purpose: records deleted in the original program, which dBase keeps in the file until the table is packed, and OLE or picture fields, which a text file can't carry. Our test set also includes a table that was damaged on purpose. It's turned down with a clear message rather than converted into something half right.

Choosing the settings

Delimiter

Comma is the standard choice, and what most databases and scripts expect. Choose Semicolon if the file is headed for a spreadsheet set up for a decimal comma, as in much of Europe; there, a comma-separated file can open with everything in one column. Tab suits tables whose text is full of commas.

Whichever you pick, numbers keep a decimal point, as in 1250.50. If your spreadsheet expects 1250,50, the workbook route, DBF to Excel, sidesteps the question, because numbers arrive as numbers.

If the CSV may go back into a DBF later, CSV to DBF works out a semicolon file's decimal mark from its numbers. The one case it can't settle is a file whose only numbers with a point look like 1.250. It reads those as 1250 and says so in the notes, which is why the comma is the safest choice for a round trip.

Text encoding

Older tables store each letter as a single byte, and the same byte means a different letter in each code page. With From the file, we read the table's language driver byte, which names its code page. We recognise 55 of them, Chinese, Japanese, Korean and Thai included. If the byte is empty or unknown, we check whether the text is valid UTF-8, and otherwise assume Western (Windows-1252). The notes always say which of these happened.

When letters still look wrong, the pattern usually points to the right choice:

• "é" where you expect "é": the text is UTF-8. Choose UTF-8.
• Symbols such as „ ‚ … ‡ where accented letters belong: the table probably comes from DOS. Try Western DOS (850), then US DOS (437), or Cyrillic DOS (866) for Russian.
• Wrong letters in Polish, Czech or Hungarian: Central European (Windows-1250). Russian or Bulgarian: Cyrillic (Windows-1251). Turkish: Windows-1254.

Change the setting, then add the file again. The new row uses the new encoding, so you can compare the two.

What the CSV looks like

The file is UTF-8 with a byte order mark, so Excel recognises accented letters, and each record ends with CR LF. A value is quoted only when it needs to be: when it contains the delimiter, a quote or a line break, as multi-line memo text often does.

• Dates are written as 2024-01-31, and date-times as 2024-01-31 14:30:00.
• True/false fields become TRUE and FALSE. A field that was never set stays empty.
• Numbers keep the decimals their field declares, so 3 in a two-decimal field is written 3.00.
• Long whole numbers, such as 18-digit IDs, are written digit for digit, never rounded.
• The spaces that pad text fields to their full width are removed.

If something goes wrong

• "This file looks damaged…" The first byte of a DBF names its version, and its field list has to add up to the record length. When either check fails, the file is usually something else renamed to .dbf, or a copy that broke off part way.
• The notes say the file ends early. The header promises more records than the file holds, which usually means an interrupted copy or download. Every complete record is converted; the original copy may have the rest.
• A memo column is empty. Its text lives in the .dbt or .fpt file. Select the table and its memo file together; their names must match apart from the extension, in any letter case. If the notes say memo values were not found, the memo file probably belongs to another copy of the table, such as an older backup.
• A column is missing. General and picture fields (G and P) hold OLE objects and images, which a CSV can't carry. The notes name each one.
• The rows are in a different order than in your program. We write records in the order they're stored. Programs often show them sorted through an index file (.cdx, .mdx, .ndx or .ntx), which isn't needed here. Sort the CSV afterwards.
• "Something unexpected stopped this conversion." Try again, and reload the page if it happens twice. Rows are written as they're read, so a big table needs far less memory here than as a workbook, though the table and its CSV both have to fit in the tab. A file over 200 MB asks before it starts, and files over 1 GB stop with a message: a limit we set so the tab doesn't run out of memory.

Try it with a sample file

No DBF at hand? Download sample-world-countries.dbf (519 KB), the attribute table of Natural Earth's world map: one row per country, with names in Arabic, Chinese, Greek, Hindi and many other languages. Drop it above. Its header doesn't name an encoding, so the notes will say UTF-8 was detected, and every name comes through intact.

Public domain. Source and details: sample file notes.