33 releases (21 stable)

2.0.3 Mar 23, 2024
2.0.2 May 7, 2023
2.0.1 Sep 30, 2022
1.4.1 Jul 30, 2022
0.6.2 Feb 27, 2016

#72 in Parser implementations

Download history 471/week @ 2024-02-15 549/week @ 2024-02-22 473/week @ 2024-02-29 968/week @ 2024-03-07 758/week @ 2024-03-14 1404/week @ 2024-03-21 859/week @ 2024-03-28 610/week @ 2024-04-04 806/week @ 2024-04-11 925/week @ 2024-04-18 20610/week @ 2024-04-25 22747/week @ 2024-05-02 17426/week @ 2024-05-09 26422/week @ 2024-05-16 26155/week @ 2024-05-23 24077/week @ 2024-05-30

99,255 downloads per month
Used in texlab




A library and command-line client for parsing and comparing human names.

Build Status


Uses and limitations

If you are trying to programmatically put human names into a canonical format, you are of course in a state of sin. But sometimes, something is better than nothing.

If you want to let your users enter arbitrary names but try to use a given name in a salutation, or you're looking for a person's index in PubMed's list of authors for a co-authored paper, or you want to extract surnames from unstructured name entries for a frequency analysis, you might find this useful.

Most names are not unique. human_name will tell you that "J. Doe" might be "Jane Doe", but of course it might not be, and even "Jane Doe" might not be the other "Jane Doe." The comparison logic here is only useful if you have external reason to believe two names may represent the same person.

human_name will work best on Latin names - i.e., data from North or South America and/or Europe. For example, it doesn't understand surname-first formats without commas, common in East Asia: "Park Geun-hye" will be parsed as having the given name "Park", and the last name "Guen-hye". And it doesn't handle single-word names. It won't blow up on Unicode, and it handles non-ASCII punctuation and accents with some intelligence, but don't feed in Arabic or Han characters and expect better results than a naive whitespace or word-boundary split.

human_name tries to fail nicely, such that if parsing fails, either it will do so explicitly, returning nothing, or at least, calling display_full on the result will return the input, modulo whitespace. But there are no guarantees.

human_name tries aggressively to treat strings as names, which makes it definitely not suitable for extracting names from a larger piece of text (although it will strip titles, nicknames, etc, from a name field.)

Because the goals of this library include both name comparison and memory efficiency, parsed names are Unicode NFKD-normalized and capitalized in a conventional way (handling "Mc" and a few other edge cases), and the raw input is not preserved.

From Rust code

use human_name::Name;

let jane_doe = Name::parse("Jane Doe").unwrap();
let john_doe = Name::parse("John Doe").unwrap();
let j_doe = Name::parse("Doe, J.").unwrap();


let oscar = Name::parse("MR OSCAR DE LA HOYA JR").unwrap();
assert_eq!(Some("Oscar"), oscar.given_name());
assert_eq!("de la Hoya", oscar.surname());
assert_eq!(Some("Jr."), oscar.generational_suffix());
assert_eq!(Some("Mr."), oscar.honorific_prefix());
assert_eq!("Oscar de la Hoya, Jr.", oscar.display_full());


See the docs for details.

From the command line

There are two modes, "parse" and "eq". The mode is passed as the first argument. You can pass input as subsequent arguments:

$ human_name parse "Jane Doe"

$ human_name parse "MR OSCAR DE LA HOYA JR"
{"first_initial":"O","given_name":"Oscar","generational_suffix":"Jr.","surname":"de la Hoya"}

$ human_name eq "Jane Doe" "Jane M. Doe"
$ echo $?

$ human_name eq "Jane M. Doe" "Jane H. Doe"
$ echo $?

Or, with the second argument "-", you can pass input on stdin. For example, to find the most common surnames in a file of newline-delimited names:

$ human_name parse - < names.txt | jq .surname | sort | uniq -c | sort -nr | head -n3
111 "Zhang"
109 "Li"
106 "Wang"

To find all the possible "J. Smith"s in a file of newline-delimited names:

$ human_name eq - "J Smith" < names.text
Smith, Jason A.
Jay Smith

Optional Features

The following features are optional and off by default:


Implements Eq for Name using consistent_with, and Hash using surname_hash.

Optional because both of these implementations are questionable for general-purpose use: this Eq is not transitive, and Hash is collision-prone. See docs for details.


Implements serialization for Name using serde. This serialization format is intended to allow programs not using human_name to see the parse results; deserialization isn't implemented because when round-tripping is desired, just using display_full to serialize as a string and then parse to deserialize should produce a more compact and reasonably performant result.

Bindings in other languages

Ruby bindings using the ffi gem are available at github.com/djudd/human-name-rb

Python bindings using the ctypes module are available at github.com/djudd/human-name-py

Elixir bindings (third party) are available at https://github.com/amokan/human_name


This library follows semver with respect to its programmatic API. Changes to the parsing and consistency-checking heuristics are not considered breaking and will typically come with a minor version bump.


As of version 2.0, the fast path (roughly, two space-separated, titlecase ASCII words) for name parsing takes ~200ns on my M1 Macbook and does not heap-allocate. Comparing two simple names for consistency takes ~100ns if the surname hash matches, and ~1ns otherwise. Pathological cases for either can take an order of magnitude longer.


Contributions, feature requests and bug reports are welcome. Please open a GitHub issue or pull request, include as much helpful context as you can, and we'll figure it out from there. As this is a small personal project, you might not get an immediate response, but I'll follow up as soon as I'm able.

Conversations related to this project must follow the Contributor Covenent's Code of Conduct. Please report any related concerns to the contact email in Cargo.toml.


Inspiration, heuristics, and test cases were taken from:


Apache 2.0 - see LICENSE.


~165K SLoC