Open-source METAR and TAF parsers compared: language, licence and upkeep
Five open-source libraries were compared on 2026-11-19: metar-taf-parser for TypeScript, python-metar-taf-parser, python-metar, avwx-engine and a Java METAR and TAF parser. Four use the MIT license, python-metar uses a BSD-style text that GitHub does not classify, and python-metar handles METAR only according to its description.
Key points
- metar-taf-parser (TypeScript, MIT) is the library this site uses, at version 9.1.3, released on 2026-03-17.
- python-metar-taf-parser, avwx-engine and the Java MetarParser are all MIT-licensed and were pushed to within the last month before the check.
- python-metar is the most starred of the five but describes itself as parsing METAR reports only.
- Stars and push dates measure activity, not correctness, and they change daily.
- No library was tested against a common set of reports, so no accuracy ranking is offered.

Why compare parsers
A METAR or TAF is a compact string, and decoding it by hand is a good way to learn the format. Software that decodes thousands of reports needs a parser, and writing one from scratch is harder than it looks, because the code has international and national variants. The US versus ICAO comparison lists differences in units, visibility and remarks that a parser has to handle. This page compares five open-source libraries by language, licence, apparent coverage and maintenance, using the public repositories as read on 2026-11-19.
The five libraries
The figures below come from the GitHub repository pages and their interface at the consultation date. Stars and push dates change daily.
| Library | Language | Licence | Stars | Last push | Latest release |
|---|---|---|---|---|---|
| metar-taf-parser (aeharding) | TypeScript | MIT | 65 | 2026-03-17 | v9.1.3, 2026-03-17 |
| python-metar-taf-parser (mivek) | Python | MIT | 67 | 2026-11-16 | v1.13.1, 2026-05-31 |
| python-metar | Python | Not classified by GitHub | 310 | 2026-10-26 | v2.0.2, 2026-10-26 |
| avwx-engine | Python | MIT | 126 | 2026-11-12 | 1.9.11, 2026-11-10 |
| MetarParser (mivek) | Java | MIT | 52 | 2026-11-01 | Not read |
None of the repositories was archived at the time of the check.
metar-taf-parser, for JavaScript and TypeScript
The repository describes itself as a parser for meteorological aerodrome reports and terminal aerodrome forecasts in TypeScript. It is the library this site depends on, declared at version 9.1.3, which matches the latest release tag. The last push and the last release both date from 2026-03-17, which is the oldest of the five. An older push is not a defect: a stable parser may simply need fewer changes. It does mean that any report format change after that date has not been checked here.
python-metar-taf-parser
This Python library is described as a program parsing METAR and TAF. It is MIT-licensed and was pushed on 2026-11-16, with v1.13.1 as latest release on 2026-05-31. The same GitHub account also hosts the Java MetarParser listed below.
python-metar
The python-metar package is the most starred of the five, with 310 stars, and its latest tag, v2.0.2, was created on 2026-10-26. Its description reads "a python package that parses coded METAR weather reports", with no mention of TAF. GitHub reports the license as NOASSERTION, meaning its detector did not match a standard license. The license file contains BSD-style text, so a project that depends on it should read the file itself instead of relying on the label.
avwx-engine
The avwx-engine repository describes an aviation weather parsing engine for METAR and TAF. It is MIT-licensed, has 126 stars, and its latest release, 1.9.11, was published on 2026-11-10. The repository description covers parsing only. How it fetches data, and which sources it reads, was not examined here. The data sources page covers where the raw text comes from.
MetarParser for Java
The Java library, also by mivek, is described as a program decoding METAR and TAF. It is MIT-licensed, has 52 stars and was pushed on 2026-11-01. Its release list was not read, so no version is given.
What the table does not show
The table shows activity and licensing, not correctness. Three practical differences matter more than a star count:
- Coverage of national variants. US remarks such as the sea-level pressure group and the temperature group differ from ICAO practice. A library may decode the main body but leave remarks as plain text. See the remarks page.
- Output model. Some libraries return a structured object with units, others return strings. Units vary by country, as the units page shows.
- Handling of invalid input. Real reports contain typos, missing groups and NIL reports. A strict parser may throw an error where a tolerant one returns partial output.
Choosing by situation
A web front end in TypeScript has an obvious candidate in metar-taf-parser. A data pipeline in Python can choose among three libraries, and the licence difference may decide: MIT is a short permissive text, while python-metar needs its license file read. A JVM service has the Java library. A project that must read IWXXM XML instead of text needs a different tool, since the libraries above are described as working on text. The IWXXM page covers that format.
A schematic report
A parser takes one text report and returns a structured record. The following is a conceptual outline, not the output of a specific library:
input: one raw report string
output: report type
station
observation or issue time
wind: direction, speed, gust, unit
visibility: value, unit
weather: list of codes
clouds: list of amount and base height
temperature and dew point
pressure and unit
remarks: kept as text or decoded
Each library names these fields differently, and the library documentation is the only reliable guide to its own names. A field that a library does not decode usually appears as an empty value or as raw text, so a project should check which groups matter for its own use before choosing, and should keep the raw report beside the parsed record so that a later parser fix can be applied again without a new download.
In real reports
The same string can decode differently across libraries when it includes uncommon groups. Groups worth testing include a variable wind direction range, a runway visual range group, a trend forecast, a TAF probability group and a US remarks section. The glossary of abbreviations lists contractions that a parser may leave undecoded.
Method and limits
- Repository facts were read through the GitHub interface on 2026-11-19: language, license label, star count, push date and latest release. Release information for the Java library was not read.
- No code was run and no library was tested on sample reports. Coverage statements rely on the one-line repository descriptions, not on documentation or source code.
- The license of python-metar is reported by GitHub as unclassified. The text was described as BSD-style from an earlier reading of the file and is not reproduced here.
- Other libraries exist and were not reviewed. The selection is not a ranking.
- Star counts and dates change quickly, and the table is valid for the consultation date only.
Training and reference only. For flight planning, always use official briefing sources (your national weather service, AIS or approved provider).
Frequently asked questions
Which parser does this site use?
The site depends on the metar-taf-parser package for JavaScript and TypeScript, declared as version 9.1.3. The upstream repository tags v9.1.3 as its latest release, published on 2026-03-17. The same package is available to any project that runs on Node.js.
Do these libraries decode both METAR and TAF?
Based on the repository descriptions, four of the five handle both METAR and TAF: metar-taf-parser, python-metar-taf-parser, avwx-engine and the Java MetarParser. The python-metar repository describes itself as a package that parses coded METAR reports and does not mention TAF.
How should a library be tested before use?
Run it on a sample of real reports that includes edge cases: variable wind, 9999 and fractional visibility, CAVOK, RVR groups, remarks, AUTO and COR modifiers, and TAF with BECMG, TEMPO and PROB groups. Compare the output with a manual decoding. Several decoders may disagree on rarely used groups.
Is a high star count a sign of quality?
Not by itself. Stars show interest and age. A better check is the date of the last release, the number of open issues, the test suite and whether the library covers the national variants of reports that matter to the project, such as US remarks or ICAO runway state groups.
Sources
- GitHub, aeharding (2026). metar-taf-parser repository (TypeScript)
- GitHub, mivek (2026). python-metar-taf-parser repository
- GitHub, python-metar (2026). python-metar repository
- GitHub, avwx-rest (2026). avwx-engine repository
- GitHub, mivek (2026). MetarParser repository (Java)