crossref-client - An async Rust client for the Crossref REST API
Fork notice.
crossref-clientis a hard fork of MattsSe/crossref-rs, which is no longer maintained. It has diverged substantially — the client is async, targets Rust 2024, and carries breaking changes to the query and response types. Changes here are not upstreamed. The original work is retained under its MIT/Apache-2.0 licence; see LICENSE-MIT.Upstream's design is in turn inspired by sckott/habanero.
The Crossref client provides methods matching the Crossref API routes:
works-/worksroutemembers-/membersrouteprefixes-/prefixesroutefunders-/fundersroutejournals-/journalsroutelicenses-/licensesroutetypes-/typesroutestyles-/stylesroute, the CSL styles a citation can be rendered inagency-/works/{doi}/agencyget DOI minting agencytransform-/works/{doi}/transform, content negotiation
The client paces itself against the rate limit crossref reports on every
response and retries a 429, so a Crossref -- and every clone of it, which
shares the limiter -- stays inside the budget it was granted.
Install
As a library
cargo add crossref-client
As a command line tool
Every release carries a prebuilt crossref binary, so this needs no rust
toolchain. Each archive holds the binary and the two licences, and is named
after the platform alone, so these commands always fetch the newest release.
Linux and macOS
# x86_64-unknown-linux-gnu | aarch64-apple-darwin (apple silicon) | x86_64-apple-darwin (intel)
target=x86_64-unknown-linux-gnu
curl -fsSL "https://github.com/EmilLindfors/crossref-client/releases/latest/download/crossref-$target.tar.gz" | tar xz
sudo install "crossref-$target/crossref" /usr/local/bin/
The binary is unsigned. Downloaded with curl it runs as it is; downloaded
through a browser, macOS quarantines it until
xattr -d com.apple.quarantine /usr/local/bin/crossref.
Windows (PowerShell)
$target = "x86_64-pc-windows-msvc"
Invoke-WebRequest "https://github.com/EmilLindfors/crossref-client/releases/latest/download/crossref-$target.zip" -OutFile crossref.zip
Expand-Archive crossref.zip -DestinationPath .
# then move crossref-$target\crossref.exe somewhere on your PATH
From source, which needs rust 1.85 or newer:
cargo install crossref-client --features cli
Either way, crossref --version says which release you ended up with. What
the binary can do is under Command Line Application.
Usage
Create a Crossref client:
let client = builder.build?;
If you have an Authorization token for Crossref's Plus service:
let client = builder
.token
.build?;
Encouraged to use the The Polite Pool:
Good manners = more reliable service
Anonymous clients share a pool limited to one request per second and will start
returning 429. Passing an email moves you to the polite pool, which currently
allows three per second. To get into it, include an email address
let client = builder
.polite
.token
.build?;
Constructing Queries
Not all components support queries and there are custom available parameters for each route that supports querying.
For each resource components that supports querying there exist a Query struct: WorksQuery, MembersQuery, FundersQuery. The WorksQuery also differs from the others by supporting deep paging with cursors and field queries.
otherwise creating queries works the same for all resource components:
let query = new
// field queries supported for `Works`
.field_query
// filters are specific for each resource component
.filter
.order
.sort;
Note that sort, order, facet, select and sample are /works-only:
/funders, /members, /journals and /licenses answer them with a 400,
so FundersQuery and the rest offer terms, paging and -- where the route takes
one -- a filter, and nothing that cannot be sent.
Two more things a query cannot ask for, for reasons on crossref's side:
- A filter value cannot contain a
,. Crossref splits thefilterparameter on it after percent-decoding, socontainer-title:Ecology, Evolutionarrives as one filter plus another calledEvolution, and neither the encoded nor the literal form survives. Such a query is refused withError::UnsendableFilterValuerather than sent to be misread. selectalways includesDOI, whether or not it was asked for. It is the one field aWorkrequires, so a page selected without it would come back as works that cannot be deserialized at all.
Other formats
Crossref will re-serialize a registered work, so a DOI can be pulled out as
BibTeX, RIS, RDF or a citation formatted in any of the ~2 900
CSL styles without going through Work:
let bibtex = client.transform.await?;
let citation = client.transform.await?;
Examples
examples/peer_review.rsreconstructs a paper's open peer review history from the reviews crossref registers against it.examples/check_pool.rsreports which rate-limit pool a client lands in.
Get Records
See this table for a detailed overview of the major components.
There are 3 different targets:
- standalone resource components:
/works,/members,funders,prefixes,typesthat return a list list of the corresponding items and can be specified with queries - Resource component with identifiers:
/works/{doi}?<query>,/members/{member_id}?<query>, etc. that returns a single item if found. - combined with the
worksroute: The works component can be appended to other resources:/members/{member_id}/works?<query>etc. that returns a list of matchingWorkitems asWorkList.
This resembles in the enums of the resource components, eg. for Members:
Examples
All options are supported by the client:
Query Single Item by DOI or ID
Analogous methods exist for all resource components
let work = client.work.await?;
let agency = client.work_agency.await?;
let funder = client.funder.await?;
let member = client.member.await?;
Query
let query = new;
// one page of the matching results
let works = client.works.await?;
Alternatively insert a free form query term directly
let works = client.works.await?;
Combining Routes with the Works route
For each resource component other than Works there exist methods to append a WorksQuery with the ID option /members/{member_id}/works?<query>?
use crossref_client::*;
async fn run() -> Result<()> {
let client = Crossref::builder().build()?;
let works = client.member_works(WorksQuery::new("machine learning")
.sort(Sort::Score).into_ident("member_id")).await?;
Ok(())
}
This would be the same as using the [Crossref::works] method by supplying the combined type
use *;
async
** Deep paging for Works **
Deep paging results
Deep paging is supported for all queries, that return a list of Work, WorkList.
This function returns a new iterator over pages of Work, which is returned as bulk of items as a WorkList by crossref.
Usually a single page WorkList contains 20 items.
Example
Iterate over all Works linked to search term Machine Learning
use ;
async
Which can be simplified to
use ;
async
Iterate over all the pages (WorkList) of the funder with id funder id by using a combined query.
A single WorkList usually holds 20 Work items.
use ;
async
Iterate over all Work items of a specfic funder directly.
use ;
async
Command Line Application
A prebuilt binary per platform, or cargo install crossref-client --features cli
-- see Install.
Usage
Top level subcommands
Usage: crossref <COMMAND>
Commands:
works Query crossref works
cite Resolve bibtex citation keys, e.g. `@LindforsJakobsen2022`, to the works they cite
funders Query crossref funders
members Query crossref members
journals Query crossref journals
licenses List the licenses crossref works are published under
prefixes Query crossref prefixes
types Query crossref types
transform Re-serialize a work into another format through content negotiation
styles List the CSL styles a citation can be rendered in
help Print this message or the help of the given subcommand(s)
Options:
-h, --help Print help
-V, --version Print version
The works subcommand
Usage: crossref works [OPTIONS] [COMMAND]
Commands:
member Get Works of a specific Member
funder Get Works of a specific Funder
journal Get Works of a specific Journal
prefix Get Works of a specific Prefix
type Get Works of a specific Type
Options:
-i, --id <ID> The DOI of a single work. Omit to search by query terms
--user-agent <USER_AGENT> The user agent to use for the crossref client
-d, --deep-page Enable deep paging. If a limit is set, then the limit takes priority
--token <TOKEN> The token to use for the crossref client
--polite <POLITE> The email to use to get into crossref's polite pool
-q, --query <QUERY_TERMS> The free form terms for the query
-l, --limit <LIMIT> limit the amount of results
-o, --output <OUTPUT> output path where the results shall be stored
-a, --append if the output file already exists, append instead of overwriting the file
--offset <OFFSET> Sets an offset where crossref begins to retrieve items
--query-bibliographic <CITATION> Match a whole reference against titles, authors, ISSNs and publication years at once. Crossref's own citation look up
--query-title <TERM> Match against a work's title, including its subtitle
--query-author <TERM> Match against author given and family names
--query-field <FIELD=TERM> Match against any other field, e.g. `--query-field container-title=Nature`. Repeat for more fields; an unknown field lists every one crossref takes
--filter <NAME[:VALUE]> Narrow the results, e.g. `--filter from-pub-date:2020-01-01` or `--filter has-abstract`. Repeat to narrow further; every filter is ANDed
--sort <SORT> How to sort the results, such as updated, indexed, published, issued
--order <ORDER> How to order the results: asc or desc
--sample <SAMPLE> Request random works. Crossref ignores every other option when set
The cite subcommand
Usage: crossref cite [OPTIONS] <KEY>...
Arguments:
<KEY>... The citation keys, with or without their leading `@`
Options:
--candidates <CANDIDATES> How many works to weigh against the key per attempt [default: 20]
--year-window <YEARS> How many years either side of the key's year still count as the same publication [default: 1]
--spellings <SPELLINGS> How many accented spellings to guess at when the key as written finds nothing [default: 12]
--user-agent <USER_AGENT> The user agent to use for the crossref client
--token <TOKEN> The token to use for the crossref client
--polite <POLITE> The email to use to get into crossref's polite pool
-o, --output <OUTPUT> output path where the results shall be stored
-a, --append if the output file already exists, append instead of overwriting the file
Examples
Be polite: pass your email so requests are routed to crossref's polite pool
(anonymous requests are limited to roughly 1 request/second and return 429).
crossref works --polite you@example.com --limit 10 --query "machine learning"
Get a single work by DOI
crossref works --id "10.1037/0003-066X.59.1.29"
Store the output in a file instead of printing to stdout
crossref works --id "10.1037/0003-066X.59.1.29" -o output.json
This works for every subcommand
crossref <works|journals|members|prefixes|types> --id "10.1037/0003-066X.59.1.29" -o output.json
Query with paging
crossref <works|funders|members> --query "machine learning" --limit 10 --offset 200
Sorting and ordering are works only; the other routes answer them with a 400.
crossref works --query "machine learning" --limit 10 --sort issued --order asc
Look a reference up by its citation. --query-bibliographic is crossref's own
reference matching: it reads titles, authors, ISSNs and publication years out of
the string together, and finds the work where the same words split across
--query would not.
crossref works --limit 1 \
--query-bibliographic "Feynman, R. (1960). There's Plenty of Room at the Bottom. Engineering and Science, 23(5)."
Match single fields instead of the whole record. --query-title and
--query-author are the two that have their own flags; every other field
/works takes goes through --query-field field=term, which can be repeated.
crossref works --query-title "room at the bottom" --query-author feynman
crossref works --query-author feynman --query-field container-title="Engineering and Science"
A field can only be matched against one term, so asking for the same one twice is an error rather than a silently dropped flag.
Narrow what comes back with --filter, which takes any of the 90 filters
/works accepts, as name for the ones that ask whether a record has
something and name:value for the ones that ask what it is. Repeat the flag to
narrow further; crossref ANDs them.
crossref works --query-title salmon \
--filter from-pub-date:2023-01-01 --filter until-pub-date:2023-12-31 \
--filter has-abstract --filter type:journal-article
The name and the value are both read before anything is sent, so a misspelled
filter lists the ones that exist and --filter from-pub-date:2020 says that a
year is not a date, rather than crossref answering with a 400 or, worse, an
unnarrowed result set.
funders and members take --filter too, each narrowing by its own
vocabulary -- location belongs to /funders alone, and /works answers it
with a 400. journals and licenses narrow by nothing, so they offer no
such flag.
crossref funders --filter location:Norway
crossref members --filter prefix:10.1016 --filter current-doi-count:1000
A filter belonging to another route is refused with the ones this route does take:
$ crossref funders --filter has-abstract
error: invalid value 'has-abstract' for '--filter <NAME[:VALUE]>':
`has-abstract` is not a filter this route takes. Try one of: location
Resolving citation keys
cite takes the bibtex keys a bibliography is written in and finds the works
they stand for.
crossref cite @LindforsJakobsen2022 @Hopp_Coffay_Lindfors_2023
A key carries no title, only surnames and a year, so crossref cannot be trusted
to have returned the right work -- it answers every query with something. Each
candidate is therefore checked against the key before it is reported: the key's
surnames have to be credited, in the order the key gives them, starting at the
first author, and the year has to be within --year-window (one year either
way by default, since a work published online in one year and in an issue the
next is cited by both).
The verdict says how much the answer is worth.
| verdict | what it means |
|---|---|
matched |
one work, and the key vouches for it. doi and the whole work are reported |
ambiguous |
several works the key vouches for, and nothing in the key to choose between them. All of them are listed |
unmatched |
nothing the key vouches for. The nearest misses are listed with the reason each was turned down |
So the DOI of a key is
crossref cite @LindforsJakobsen2022 | jq -r '.[] | select(.verdict == "matched") | .doi'
cite exits 0 only when every key resolved to exactly one work, so a
bibliography can be checked in a script. The report is written either way -- a
key that resolved to nothing, or to several works, is an answer rather than an
error.
# every key in a .bib, checked against crossref
grep -o '@[a-z]*{[^,]*' refs.bib | cut -d'{' -f2 | xargs crossref cite --polite you@example.com \
|| echo "some keys did not resolve"
Keys that lost their diacritics
Citation keys are written in ascii and crossref folds nothing, so
query.author=Floysand finds the works of a different person and
query.author=Fløysand is the only way to reach @FloysandEtAl2021. When the
key as written finds nothing, cite guesses the marks back one letter at a
time, likeliest first, and reports the spelling that worked as matched-as.
crossref cite @FloysandEtAl2021 # three requests: as written, as a citation, then Fløysand
Each guess costs a request, so --spellings caps how many are made (twelve by
default, 0 to make none). Verification folds both spellings together, so a key
written either way is checked against metadata written either way.
Get the works of a specific member
crossref works member 98
By default deep paging is disabled, so at most a single crossref page (20 Work items)
is returned. Pass --deep-page to page through the whole result set with a cursor.
crossref works --deep-page --query "machine learning" -o all.json
Where the api's own behaviour is written down
When a route answers something this crate did not expect, these are the places to look before assuming the bug is here:
- The swagger UI — the routes and their parameters, as the api reports them today.
CrossRef/rest-api-doc— the prose documentation, including the work record format every response type here is modelled on.- Crossref's public issue board
— the
CRproject, open to read without an account. Api bugs, indexing gaps and in-flight changes are tracked here, so a filter that stopped matching or a field that changed shape usually has a ticket before it has an explanation anywhere else.
The live suite (cargo test --test integration) asks the api the same
questions and is the fastest way to tell a change on crossref's side from a
change on this one.
License
Licensed under either of these:
- Apache License, Version 2.0, (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or https://opensource.org/licenses/MIT)