Notable technical changes to the fatcat schemas, Rust API server, Python web interface, and Python client library will be recorded here.
Intent is to follow semantic versioning, with all components kept in lock-step,
but note that everything is pre-1.0, so there aren't any actual backwards
compatibility guarantees. The API prefix (eg, /v0/
) will probably roughly
track major version number, but TBD.
See also:
- release documentation in
./extra/release_process.md
- Keep a Changelog
- Semantic Versioning
This is mostly a roll-up release of changes from the last year, as an arbitrary milestone.
No changes to API or SQL schemas.
- web: container "browse" views (eg, by volume)
- guide: more content, particularly "for publishers" and "for authors"
- 'stub' releases are not included in container preservation coverage "by year" and "by volume" charts. they are still included in overall counts.
- rust: upgraded 'macaroon' crate to upstream v0.3.0, instead of patched version
This is a minor version bump to mark the upgrade of dependencies. There are no
schema changes, and the schema version has not been incremented (is still
0.5.0
).
- expected and tested deployment environment is now Ubuntu Focal and PostgreSQL 13
- minimum rust version is now 1.41. no new features are being used, but older versions will no longer be tested or supported
- default local development database now called
fatcat_dev
, notfatcat
- various typos and spelling errors corrected (using
codespell
)
- "alert banner" mechanism for web interface
- API server detects SQL transaction failure due to read-only mode, and returns appropriate 5xx error
Small change to the API schema (and SQL schema), adding the content_scope
field to file, fileset, and webcapture entities. Because there is a SQL schema
change, bumping to version v0.5.0.
An outward-facing change is that the fatcat API server now attempts to "stabilize" array order within JSON responses by sorting elements by appropriate identifiers or other keys. In many cases, the order of things like URLs for a file, manifest entries for a fileset, or releases associated with a file, may be reversed in API responses compared to what was returned previously. They should not match what was original supplied when the entity was created.
In particular, this may cause broad discrepancies compared to historical bulk metadata exports. New bulk exports will be generated with the new ordering.
A number of content cleanups and changes are also taking place to the primary catalog (fatcat.wik), see the separate content CHANGELOG for details.
- API array order stablization, using
ORDER BY
infatcatd
. See note above.
- broad python code style updates: formatting, lint rules, and type annotations
- a number of internal refactors of metadata importers
- stopped created a small number of Datacite-specific license slugs
- stopped trying to "fix" double slashes in DOIs, in most cases
- reduced amount of metadata stored in release
extra
field in Datacite importer
content_scope
field on file, fileset, and webcapture entities- initial fileset importers
- JSON pseudo-API for reference string match/get interface
- editgroup "diff" view in web interface
- deleted deprecated
cdl_dash_dat
andwayback_static
one-time importers
Includes small API and SQL schema changes; see
./proposals/20211012_v04_schema_tweaks.md
. Old database dumps may not work
with this version of API server; migrations can be run using the diesel
command line tool to upgrade old dumps.
- reference graph views, based on fuzzy reference dataset in
refcat
andfatcat-scholar
projects, stored in elasticsearch index - savepapernow webcapture/html importer
- release schema: new
hdl
(handle) external identifier - fileset schema: new
mimetype
field for files under manifest - container schema: new
issne
andissnp
top-level fields, which are supported in lookups, as well as theissn
lookup field, which matches any of theissn*
fields - container schema: new
publication_status
field, to indicate if container is still actively publishing, discontinued, or "stub" - above API schema changes added to elasticsearch 'release' and 'container' schemas
- editor "lookup" API endpoint, for fetching editor object by username
- web
/u/{username}
redirect helper for inspecting editors, bots, etc
- viewing deleted release entities no longer result in 500 error
- mediawiki (wikipedia) OAuth logins (bytes/str bug in fatcat code)
- savepapernow ingest importer force flushes every 5 minutes; should fix a bug with many SPN ingest requests not getting imported
- added 'version' field to release elasticsearch schema (was missing)
- several release external identifier validators were not actually being run before insertion
Mostly infrastructure changes: Python, Elasticsearch, and other dependencies changed.
No changes to schema, so API remains v0.3.3.
- require Python 3.8 (upgrade from Python 3.7)
- require Elasticsearch 7.10 (upgrade from 6.x). Queries (reads) will work against ES 6.x, writes will not. 7.10 is the last upstream "open source" version of Elasticsearch, and we intend to stick with it until a community open source fork emerges
- update all Elasticsearch schemas to v03c. All at least include
doc_index_ts
, and small other changes - container Elasticsearch schema includes aggregated release preservation numbers, as well as total count. Indexing pipeline updated to (optionally) query these stats at index-time
- several improvements to web editing workflow: clearer design of editgroup web view, etc
- change web UI and labeling on release and file views to emphasize preservation status and archival access
- update Flask to 2.x, and other related dependencies
- display preservation holdings summary on container view page
Minor additions to the API schema: new external identifiers for release
entities, for doaj
, dblp
, and oai
. Database schema (SQL) not changed.
- three new release external identifiers:
doaj
,dblp
, andoai
(all article-level). These are API changes, but backwards compatible. - DOAJ release import
- dblp container and release import
- free-form "coverage search" page, allowing visualization of coverage based on elasticsearch query (web interface)
- editing of all entity types using TOML markup (via web interface)
- basic sitemap XML generation
- initial integration of fuzzycat library to prevent duplicate release entity creation at import time
- import of HTML webcaptures from sandcrawler ingest
- kafka publishing of updated work entities (transitively of release updates as part of the work), to enable work-level entity update pipeline for archive scholar index
- require Python 3.7 (upgrade from Python 3.5)
- release entity exports can now be sorted by work identifier, for easier work-level grouping and analysis
- refactored webface search code to use elasticsearch client library
- several datacite metadata import bugs
- several other bugfixes to web interface and importer code, not reported here granularly
This release was tagged retro-actively; it was the last commit before upgrading to Python 3.7.
Many small changes and tweaks to importers, web interface, etc were made in this release.
- pubmed importer
text
vs.get_text()
for HTML tags - fatcat-python-client package now works with Python 3.7, with removal of
async
keyword
- fatcat-python-client re-code-generated using openapi generator instead of swagger tooling
- minimum rust version now 1.36
- Switch from swagger-codegen to openapi-generator for python client generation
- switch python Kafka code from pykafka to confluent-kafka
- update release and container elasticsearch schemas to v03b. Release search is now over "biblio" field, allowing matches on multiple fields at the same time
- Crossref harvester using 'update-date' not 'index-date' to detect updated documents
- OpenSSL support removed from fatcatd (Rust)
#@# Added
- webface endpoints for entity view URLs with an underscore instead of slash,
as a redirect. Eg,
https://fatcat.wiki/release_asdf
=>https://fatcat.wiki/release/asdf
. A hack to make copy/paste easier. - pagination of search results in web interface
- sandcrawler daily crawling pipeline, including ingest-file importer and publishing requests to sandcrawler kafka topic
- "Save Paper Now" feature (using sandcrawler pipeline)
- Datacite DOI registrar daily harvesting and importing
- Arxiv daily harvesting, using OAI-PMH worker
- Pubmed daily harvesting, using FTP worker
- "file" entity elasticsearch schema (though pipeline not yet running continuously)
Many small web interface improvements, bug fixes, and importer additions are not explicitly noted below.
- renamed python client library
fatcat-openapi-client
and rust api-spec libraryfatcat-openapi
- updated rust (cargo) dependencies; depend on rust 1.34+ (Debian buster)
- entity view page design and backend code structure completely refactored
- compilation of
fatcat-openapi
rust crate (including examples) now actually works using 2018 edition. Many lint warnings were patched, making rust compilation cleaner with recent compiler versions.
- container "coverage" page, showing elasticsearch-generated chart of IA release archival coverage
create_auth_token
API endpoint, for generation of API tokens via web interface- API token generation in web interface
- quick web interface search (front page and top bar) now detects additional identifiers, such as ISBN-13, SHA-1, and arxiv
This version includes both a SQL migration and backwards-incompatible API changes.
edit_extra
on individual entity create was not being written to database
release_status
field (on releases) renamed torelease_stage
- moved external identifier fields on release entities into a new sub-namespace
called
ext_ids
. Eg, instead ofrelease.doi
, it's nowrelease.ext_ids.doi
. This field is always required for release entities, so fields can be accessed directly without a null/option check on theext_ids
field itself. This impactsdoi
,pmid
,pmcid
, etc. pmcid
field now accepts versioned identifiers, with a trailing dot and version number, likePMC1234.2
.- many more restrictions on external identifiers at creation time.
- API endpoints that mutate entities are now prefixed with
/endpoint/{endpoint_id}/...
, which changes theendpoint_id
parameter from optional to required (in the path). In client libraries, this also changed the order of parameters (endpoint_id
now comes first). Batch edits now only work in "auto batch" mode, and an editgroup object must be included in the body (not justextra
anddescription
fields as query parameters). - several additional editgroup/changelog/history endpoints now expand
editor
in editgroup objects by default. created
timestamp is included in editgroup objects (when GET) by default.- in client libraries,
FileEntityUrls
renamedFileUrl
, and several similar singular/plural andEntity
removed renamings. - elasticsearch release schema updated to match API schema changes
- Non-auto batch mode no longer implemented.
- release contribs may include
given_name
andsurname
fields, likecreator
objects, in addition to existingraw_name
. - add
withdrawn_status
,withdrawn_year
, andwithdrawn_date
fields to releases - added
retraction
as an allowablerelease_type
, for a publication. When used, therelease_stage
should also beretraction
. subtitle
added as a top-level release field, along withversion
andnumber
ark_id
(for ARK identifiers) andmag_id
(for Microsoft Academic Graph identifiers) added to releases (underext_id
)- allow expanding
releases
for file, fileset, and webcapture entities. Expanded release entities have their abstracts and refs hidden by default (for performance) creator_ids
in release elasticsearch schema, for lookups- new importers: arxiv, pubmed, JALC, and JSTOR. Most still need refactoring and polish
- fixed major authn/authz bug with fatcat python client which leaked API tokens between API client handles. Almost all tests/webfact/etc were potentially running with the privileged (superuser) webface-bot privileges. Yikes!
- API
get_editgroup_annotations
endpoint was requiring auth; this was a typo. Going to call this a very minor/backwards-compatible API change and not do a minor version bump for it. - DOI lookups (for release entities) were not being lower-cased as intended (DOIs are stored lower-case internally)
- ORCID OAuth login provider
- small tweaks to URL domain/rel mapping list
No API or SQL schema changes in this release. Macaroon generation and verification was broken; all non-CLI-generated tokens will need to be regenerated (eg, log-out, log-in).
- fix macaroon 'expires' caveat check (in rust), allowing actual login via OAuth/IA
- basic release, container, file webface editing
- basic editgroup annotation and control (submit/accept) in webface
- CSL/BibTeX citation endpoints
- "review bot" framework (for CI-style edit checks)
- wikipedia OAuth login provider
- expanded example entities in SQL schema
- updated rust (cargo) dependencies; depend on rust 1.32+
- many tweaks to webface and guide
First tagged release, and start of notable change tracking.