fair-cli 0.11.0__tar.gz → 0.12.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (43) hide show
  1. {fair_cli-0.11.0 → fair_cli-0.12.1}/CHANGELOG.md +133 -0
  2. {fair_cli-0.11.0 → fair_cli-0.12.1}/CITATION.cff +2 -2
  3. fair_cli-0.12.1/PKG-INFO +576 -0
  4. fair_cli-0.12.1/README.md +528 -0
  5. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/__init__.py +15 -0
  6. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/cli.py +308 -24
  7. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/exceptions.py +2 -2
  8. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/history.py +74 -12
  9. fair_cli-0.12.1/fair/registry/database.py +235 -0
  10. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/lifecycle.py +255 -16
  11. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/requests.py +76 -7
  12. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/server.py +424 -8
  13. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/storage.py +28 -4
  14. fair_cli-0.12.1/fair/registry/stores.py +391 -0
  15. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/sync.py +1057 -154
  16. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/session.py +572 -39
  17. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/staging.py +77 -7
  18. fair_cli-0.12.1/fair/templates/hist.jinja +3 -0
  19. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/user_config/__init__.py +165 -15
  20. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/user_config/globbing.py +5 -4
  21. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/user_config/validation.py +65 -0
  22. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/utilities.py +34 -0
  23. {fair_cli-0.11.0 → fair_cli-0.12.1}/pyproject.toml +1 -1
  24. fair_cli-0.11.0/PKG-INFO +0 -448
  25. fair_cli-0.11.0/README.md +0 -400
  26. fair_cli-0.11.0/fair/templates/hist.jinja +0 -3
  27. {fair_cli-0.11.0 → fair_cli-0.12.1}/LICENSE +0 -0
  28. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/common.py +0 -0
  29. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/configuration/__init__.py +0 -0
  30. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/configuration/validation.py +0 -0
  31. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/files.txt +0 -0
  32. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/identifiers.py +0 -0
  33. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/logging.py +0 -0
  34. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/register.py +0 -0
  35. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/__init__.py +0 -0
  36. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/file_formats.json +0 -0
  37. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/file_types.py +0 -0
  38. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/registry/versioning.py +0 -0
  39. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/run.py +0 -0
  40. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/templates/__init__.py +0 -0
  41. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/templates/config.jinja +0 -0
  42. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/testing.py +0 -0
  43. {fair_cli-0.11.0 → fair_cli-0.12.1}/fair/virtualenv.py +0 -0
@@ -1,3 +1,136 @@
1
+ # 2026-11-15 [v0.12.1](https://github.com/FAIRDataPipeline/FAIR-CLI/releases/tag/v0.12.1)
2
+
3
+ ## Fixed
4
+ - v0.12.0 has no binaries: the Windows build's check of the binary's version failed, though the binary ran and gave
5
+ the right version, so no binary was released for any platform. v0.12.1 releases them for Windows, macOS and Linux,
6
+ with the same code as v0.12.0.
7
+
8
+ ## Development
9
+ - The deploy workflow checks each binary's version without `tee /dev/stderr`, as Git Bash on the Windows runner has
10
+ no `/dev/stderr`.
11
+
12
+ # 2026-11-11 [v0.12.0](https://github.com/FAIRDataPipeline/FAIR-CLI/releases/tag/v0.12.0)
13
+
14
+ ## Changed behaviour
15
+ - `fair run` refuses a configuration with no `description` in `run_metadata`, or a blank one. So do `fair run --ci`
16
+ and `--rehearse`. The registry needs a description of every code run, and without one the run's API failed later.
17
+ The refusal comes before anything is registered. `fair pull` needs no description.
18
+ - `fair run` and `fair pull` check each name that a `write:` or `register:` entry would register. They refuse a name
19
+ that breaks the naming convention before registering anything, and name the entry. A namespace holds ASCII letters,
20
+ digits, `_` and `-`. A data product's name is one or more such parts joined by single `/`. A part may also hold
21
+ `.`, but not at its start or end. A `read:` entry is not checked.
22
+ - `fair registry install` without `--directory` now looks for the registry in the directory the configuration names.
23
+ With `--force` it replaces the registry there. Before, it installed in `~/.fair/registry` whatever the
24
+ configuration said. When a registry is installed already, the refusal names `fair registry upgrade`.
25
+ - `fair init --using <file>` and `fair init --ci` keep a machine's global configuration where there is one. Before,
26
+ they replaced its user, namespaces, remote, local registry and data store for every project on the machine. Their
27
+ settings are now the project's alone, and the project uses the machine's local registry. A file naming another
28
+ local registry, port or data store is refused; a project's own store is `run_metadata: write_data_store`.
29
+ - A `read:` entry's `use: cache` names your own copy of the data product's file, such as a private one the remote
30
+ cannot send. `fair pull` now pulls the entry as any other read, and copies that file into the local data store in
31
+ place of the download. It refuses a file without the hash the remote records. Before, it pulled nothing for the
32
+ entry, and `fair run` let it through unregistered. A wildcard `read:` entry with `cache` is refused.
33
+ - A `write:` pattern's version token is now bumped from the highest version of any name the pattern matches, and
34
+ every name written under the pattern takes that version. Before, a name already registered was bumped from its own
35
+ highest version, so it could be written below another name under the pattern. A `write:` entry for one name now
36
+ takes the place of a pattern's entry for that name. Two `write:` patterns matching one registered name are refused.
37
+ - `fair add` refuses a code run that records no inputs or outputs. A run records them as it finishes, so such a run
38
+ is still running, or stopped before it finished.
39
+
40
+ ## Added
41
+ - `fair add --all` stages every finished code run of this repository that the remote lacks, or holds only in part.
42
+ `--remote` names the remote, `origin` by default. A run has finished once it records inputs or outputs. It is
43
+ refused while another process holds a session on the local registry, since that may be a run in progress.
44
+ - `fair rm <uuid>...` deletes code runs from the local registry and from the project's staging. It refuses a run the
45
+ remote holds (`--local` does not ask it), and a run whose outputs another run read, unless that run is deleted too.
46
+ It asks first (`--yes` answers for you). What the runs wrote stays registered. Before, `fair rm` did nothing.
47
+ - `fair list code-runs` lists every code run, not the first ten. It shows whether each has finished, whether the
48
+ project has staged it, and how much of it `origin` holds. `--json` prints them as JSON, for a program to read.
49
+ - `fair log -n <count>` lists the newest `<count>` jobs, and `fair log --all` every one. Ten is still the default.
50
+ - `fair registry erase-unknown-data` erases the files in the local registry's data stores that the registry does not
51
+ hold. It runs only while the registry is stopped. It names each store and warns that another registry using it may
52
+ hold the files, then erases them only once the store's path is typed. `--dry-run` lists what it would erase. It keeps
53
+ any file that a run in progress could still be writing, unless given `--delete-all`. We recommend a data store for
54
+ each local registry.
55
+ - `fair registry reset` empties the local registry. It deletes every record and keeps the install. It asks first
56
+ (`--yes` answers for you). It registers the file types again, and restarts the registry if it was running. If it
57
+ fails, or is stopped by Ctrl-C, the registry keeps its records. Run in a project, it also registers the project's
58
+ user again and clears the project's staging. `--delete-files` also deletes the files the registry held in its data
59
+ stores.
60
+ - `fair registry upgrade` moves the local registry to another version: the newest release, or the one given by
61
+ `--version`. It keeps the registry's records if the new version's tables are the same. If they differ, it leaves
62
+ the registry as it was, as it does if it fails or is stopped by Ctrl-C. Given the installed version, it rebuilds the
63
+ registry. That mends one whose Python has gone.
64
+ - While `fair registry erase-unknown-data`, `reset`, `upgrade` or `install --force` runs, any other command that would
65
+ start, stop or use the registry is refused. The refusal names the command that is running, and so does
66
+ `fair registry status`.
67
+ - `fair push` and `fair pull` send the records attached to what they send. These are named components, licences,
68
+ keywords, key-values, quality-control records, code releases, and the issues on components. A record added after its
69
+ object was sent goes with the next push or pull of that object. If the other registry holds a record with other
70
+ values, the record there is kept as it is, with a warning.
71
+ - An issue is not sent again if the other registry holds it already. It is found there by its UUID, or else by its
72
+ severity and description on the same component.
73
+ - `fair push` warns of earlier code runs that the remote does not hold. Each such run wrote a pushed data product, an
74
+ input of a pushed run, or an input of another such run. The warning names the `fair add` command that would send it.
75
+
76
+ ## Fixed
77
+ - `fair push` took time growing with the square of the number of data products a code run read: about 23 hours for a
78
+ run that read 600. A push now sends or checks each data product once, and links each code run once, after the data
79
+ products are sent. A push that fails links the runs of what it sent before the failure.
80
+ - A code run that read or wrote a named part (a component) of a data product could not be pushed. The push failed
81
+ with "Failed to access component ... on remote registry". A run is now recorded on the remote as reading and
82
+ writing the parts it did. Before, a run pushed by itself was given every part of each file it used.
83
+ - When two machines ran one commit of a code repository and pushed to the same registry, the registry got two objects
84
+ for that commit. The repository's release was on only one of them. If the machines registered the repository at
85
+ different paths, the push failed instead. A push or pull now gives a run the object that the other registry holds
86
+ for its commit. If there are several, it takes the one with a release, else the oldest, as DataPipeline.jl does
87
+ locally. A release registered as a data product gets the same object.
88
+ - A push that reached two code runs the remote held already could give the second run the first's outputs. Registry
89
+ v1.5.0 refuses that, so the push failed.
90
+ - `fair push` and `fair pull` took a data product on the other registry with the same namespace, name and version to
91
+ be the same one, whatever its file or source. A push then left the remote's data product in place without a word.
92
+ A run that used it failed with "Failed to access [] on remote registry". Both commands now check every data product
93
+ they would send before sending anything. They refuse one held there with other contents, and name each side's file
94
+ or source.
95
+ - With a web proxy set in the environment (`http_proxy`, `https_proxy` or `all_proxy`), every request to the local
96
+ registry went to the proxy. Commands then failed with an error naming the registry's token or another server.
97
+ Requests to this machine now go to it directly. Requests to a remote registry or an identifier service still use
98
+ the proxy.
99
+ - `fair add` failed unless the local registry was running. It now starts the registry, and stops it afterwards if it
100
+ started it. `fair add ""` staged an empty name, and is now refused.
101
+ - A code run deleted from the local registry stayed in the project's staging. `fair add` staged it again, and
102
+ `fair push` then failed naming it. `fair status`, `fair add` and `fair push` now remove it from staging and say so.
103
+ `fair add` refuses it, and a push sends the rest.
104
+ - Shell completion of a data product for `fair add` offered nothing: it ran `fair list` and stopped. It now offers
105
+ the data products the project's staging names.
106
+ - A registry search dropped any filter set to `false` or `0`, so it found everything.
107
+ - With no `git` on the path, every command failed with GitPython's own error, even `fair --version` and `--help`.
108
+ These two now run. A command that needs `git` says so, and names `GIT_PYTHON_GIT_EXECUTABLE`.
109
+ - `fair registry install --version` took a branch only as `origin/<branch>`. It now takes the branch's name. Given a
110
+ version that the registry's repository does not have, it stopped the running registry before refusing. It now
111
+ checks the version before stopping anything.
112
+ - `fair log` stopped with "Cannot open log for job" at the first job that another repository ran with the same data
113
+ store. By default every project on a machine shares one data store. `fair log` now lists this repository's jobs,
114
+ then says how many others it skipped and which repositories ran them. It lists at most 10 jobs, not 11. If this
115
+ repository has run no jobs, it says so.
116
+ - `fair view` stopped with a Python traceback when given a job that another repository ran. It now says it found no
117
+ such job, and names `fair log`.
118
+ - `fair log` and `fair remote` printed some text in white or yellow, which is hard or impossible to read on a light
119
+ background. A job's id and date, and a remote's name, are now bold in the terminal's own colour. A remote's address
120
+ is plain.
121
+ - `fair init --using` failed on a file with a top-level `description`, as earlier documented examples had. It now
122
+ ignores the key. A registry in the file with no `data_store` is now refused naming that key, not another.
123
+ - A user whose identifier, such as an ORCID, the local registry held under another spelling of the name could not
124
+ register: `fair init` and `fair pull` stopped with "Expected to receive a URL location from registry post". The
125
+ registry's author with that identifier is now used, with a warning naming both spellings.
126
+
127
+ ## Development
128
+ - The implementations workflow no longer installs `libmemcached-dev`, which the registry has not needed since before
129
+ v1.0.0. So the Python, Java and Julia jobs no longer need Ubuntu's package archive, which GitHub's runners could
130
+ not reach for a time.
131
+ - The C++ job of the implementations workflow no longer upgrades the runner's own packages, which it does not use.
132
+ - A storage test no longer checks its DOI online. It failed when the site the DOI leads to closed the connection.
133
+
1
134
  # 2026-10-19 [v0.11.0](https://github.com/FAIRDataPipeline/FAIR-CLI/releases/tag/v0.11.0)
2
135
 
3
136
  ## Changed behaviour
@@ -26,7 +26,7 @@ authors:
26
26
  affiliation: University of Glasgow
27
27
  orcid: https://orcid.org/0000-0002-4424-9890
28
28
 
29
- date-released: 2026-10-19
29
+ date-released: 2026-11-15
30
30
  keywords:
31
31
  - FAIR Data Pipeline
32
32
  - FAIR
@@ -36,4 +36,4 @@ license: BSD-2-Clause
36
36
  message: If you use this software, please cite it using these metadata.
37
37
  repository-code: https://github.com/FAIRDataPipeline/FAIR-CLI/
38
38
  title: "The FAIR Data Pipeline command line tool"
39
- version: 0.11.0
39
+ version: 0.12.1