ISO Language Codes (639-1 and 639-2) and IETF Language Tags

Language codes in four CSV files: the two-letter ISO 639-1 codes, the three-letter ISO 639-2 codes (bibliographic and terminologic) with English and French names, and a list of the IETF language tags that have a locale file in the Unicode CLDR.

Column definitions and types are in datapackage.json.

Data

There are four files in data/.

data/language-codes.csv has the languages that have a two-letter code, sorted by that code.

Column	Meaning
alpha2	ISO 639-1 two-letter code, for example de
English	English name of the language, for example German

data/language-codes-3b2.csv has the languages that have both a three-letter bibliographic code and a two-letter code.

Column	Meaning
alpha3-b	ISO 639-2 three-letter bibliographic code, for example ger
alpha2	ISO 639-1 two-letter code, for example de
English	English name of the language

data/language-codes-full.csv has every entry of the ISO 639-2 list.

Column	Meaning
alpha3-b	ISO 639-2 three-letter bibliographic code, for example ger
alpha3-t	ISO 639-2 three-letter terminologic code, for example deu. Empty when the language has no separate terminologic code
alpha2	ISO 639-1 two-letter code. Empty when the language has none
English	English name of the language
French	French name of the language

Every language has a bibliographic code, but only a few have a terminologic code. For example, German (Part 1: de) has two codes in Part 2: ger (bibliographic) and deu (terminologic). English has only one code in Part 2: eng. See how the parts of ISO 639 relate in Wikipedia.

data/ietf-language-tags.csv has one row for each locale file in the Unicode CLDR core data. It does not list every valid language tag.

Column	Meaning
lang	Language tag made from the file name, with - instead of _, for example az-Latn-AZ. The file root.xml gives root
langType	Language code, 2 or 3 letters (for example aa, agq), or root
territory	Country code (for example DJ) or CLDR region code (for example 001 for the world, 150 for Europe). Empty when the file has no territory
revGenDate	Date the file was generated. Always empty, see below
defs	Number of top-level sections in the file, not counting its identity section
dftLang	1 if CLDR lists the locale as the default content of its language (for example af-ZA for Afrikaans), 0 otherwise
file	Name of the CLDR locale file, for example aa_DJ.xml

Known limitations

• Names in English and French can hold several names separated by ; (for example Adyghe; Adygei). Use the codes as keys, not the names.
• The special codes mul, und, mis and zxx are in the full file, and so is one row with the code range qaa-qtz ("Reserved for local use"). That row is not a normal three-letter code.
• Nearly every French name in language-codes-full.csv ends with a carriage return (\r) inside the quotes. Remove it after reading: df["French"] = df["French"].str.rstrip("\r").
• territory in ietf-language-tags.csv can be NA, the code for Namibia (for example en-NA). pandas reads NA as a missing value by default. The two-letter language code na (Nauru) in the other files is not affected, because pandas only treats upper case NA as missing.
• territory can also be a three-digit CLDR code with leading zeros (001). Read it as text.
• With pandas, read the files as text and keep empty cells empty: pd.read_csv("data/ietf-language-tags.csv", dtype=str, keep_default_na=False). This keeps NA, 001 and the empty cells as they are in the file. Without it, NA becomes a missing value.
• dftLang comes from CLDR's list of default content. A few entries in that list have no locale file in the CLDR core download, so they have no row here.
• revGenDate is always empty, because the current CLDR locale files carry no generation date.
• alpha3-t and alpha2 in the full file are empty for many rows. alpha3-t is empty unless the language has a terminologic code.

Data sources

• ISO 639-2 code list, published by the Library of Congress, the ISO 639-2 Registration Authority. It provides the codes and the English and French names in the three language-codes files. Licence: unclear (see License).
• Unicode Common Locale Data Repository (CLDR), downloaded as core.zip. It provides the locale file names and a few attributes in ietf-language-tags.csv. Licence: Unicode License v3. The notice is in ATTRIBUTION.txt.

Changes made to the source data: language-codes-full.csv is the ISO 639-2 list split on its field separator into columns, with the names unchanged. language-codes.csv keeps only the languages that have a two-letter code, sorted by that code. language-codes-3b2.csv keeps only the languages that have both a bibliographic three-letter code and a two-letter code. ietf-language-tags.csv has one row for each CLDR locale file: tag, language, territory, generation date, number of definitions, default-language flag and file name. No CLDR names or translations are copied.

Not endorsed by, or affiliated with, the Library of Congress, Unicode, Inc. or ISO.

Preparation

The scripts need bash, curl, awk, GNU sed, wget, unzip and php. The data is rebuilt in the same steps as the GitHub workflow, from the root of the repository:

wget https://www.unicode.org/Public/cldr/latest/core.zip
unzip core.zip
mv common/main/ ./
./scripts/language-codes.sh -f
php scripts/ietf-lanGen.php

scripts/language-codes.sh downloads the Library of Congress list and writes the three language-codes files. scripts/ietf-lanGen.php reads the locale files in main/ and writes data/ietf-language-tags.csv. It also downloads a list of default locales from the unicode-cldr/cldr-core repository to set dftLang. The workflow then deletes the CLDR files it unpacked.

A GitHub workflow runs this every Monday at 02:00 UTC, and on every push to main. It commits the data files if they changed.

License

The licence of the ISO 639-2 list is unclear. What we checked:

• The list file and the code list page of the Library of Congress carry no copyright or licence notice, and neither do the Registration Authority page or the FAQ.
• The Library of Congress legal page says that it generally does not own the rights to the materials it shows and that you should decide for yourself whether an item is protected. It does not mention this list.
• ISO states that it "allows free-of-charge use of its country, currency and language codes from ISO 3166, ISO 4217 and ISO 639" (read on an archived copy of the ISO page, December 2024; ISO's current page could not be read). ISO also says that reproducing its materials needs its written permission.

We could not establish whether the full list, with the English and French names, may be republished or used commercially. If you plan commercial use, check the terms of the Library of Congress and ISO first.

data/ietf-language-tags.csv is derived from the Unicode CLDR, which is under the Unicode License v3. That licence allows use, copying and sale if the copyright and permission notice stays with the data. The notice is in ATTRIBUTION.txt.

The maintainers place only their own work (the selection and formatting of the data) under the Public Domain Dedication and License (PDDL). This does not cover the ISO 639-2 list or the CLDR. See also ATTRIBUTION.txt.