- Query/List/summarize all packages
- Query/List/summarize all package versions
- filter package and version lists by any value in the result json.
- Both
include
andexclude
filtering - Sort by any value in the result json.
- full regex matching capabilities
- result-set slicing python list slicing
- Delete Package Versions
- return the summary json back to the
output
option of the Action. - extract a specific value from the json response into the
output
option of the Action. - Better Feedback of processing during a long running operation.
See the Common Tasks
section below for a few useful examples.
- name: Org owned packages
uses: shaneapowell/ghaction-package-admin@v0
with:
ghtoken: ${{ github.token }}
operation: deletePackageVersions
org: ${{ github.repository_owner }}
package_type: container
package_name: <your package name>
include: metadata.container.tags[*] .*-latest
exclude: metadata.container.tags[*] .*-develop
sort_by: updated_at
revese: false
summary: true
debug: false
slice: 5 __NONE__
dryrun: ${{ inputs.dryrun }}
- name: User owned packages
uses: shaneapowell/ghaction-package-admin@v0
with:
ghtoken: ${{ github.token }}
operation: deletePackageVersions
user: ${{ github.repository_owner }}
package_type: container
package_name: <your package name>
include: metadata.container.tags[*] .*-latest
exclude: metadata.container.tags[*] .*-develop
sort_by: updated_at
revese: false
summary: true
debug: false
slice: 5 __NONE__
dryrun: ${{ inputs.dryrun }}
__NONE__
is a special string that is used to pass in a blank/null as a parameter value. From the command line, you can simply leave off the option, and it will be ignored. This special string provides a means for the yml based action to be able to pass through all expected parameters, and still allow for a parameter to specifically ignored. In other cases, it might be needed to pass a null/None
value to an option. This is especially true in the use of the --slice
option.
Takes a single value. The operation to perform. Must be one of the following strings. More options To be added in future releases.
listPackages
- Lists the packages owned by the given user/org.listPackageVersions
- Lists the package versions owned by the given user/org.- must also provide
--package_name
option.
- must also provide
deletePackageVersions
- Given the result of the filters and sort, delete the found items.- must also provide
--package_name
option.
- must also provide
The PAT
GitHub token. The permissions of that token must be sufficient to perform the actions in question. see the FAQ for examples of errors and possible solutions.
If this is an Organization owned package. Provide this --org
and do NOT provide the --user
option.
If this is an User owned package. Provide the --user
and do NOT provide the --org
option.
One of the GitHub package type codes. If you are using ghcr.io, you'll want to use the container
type.
See list packages for an organization for all github types
npm
maven
rubygems
docker
nuget
container
The name of your package.
The filter to include records in the result list. see Include/Exclude Filters
section below.
This filter will find matches in the result set and keep them in the list. Removing those that don't match.
An include filter is made up of 2 string values. a json-path
and a regex
separated by white space.
--include metadata.container.tag[*] develop-.*
--include created_at 2024-03-.*
Same as the --include
filters, except this filter excludes it's matches. You can combine --include
and --exclude
in a single run.
--exclude metadata.container.tag[*] latest
--exclude metadata.container.tag[*] .+-latest
--exclude metadata.container.tag[*] v1.*
Sort the result set by the given json-path
identified field. The default is a natural string sort of the values in this field. Sorting is done after filtering is complete.
--sort_by updated_at
--sort_by metadata.package_type
--sort_by id
Reverse the sort_by
result. true
or false
--reverse true
Perform a python list slice
operation on the result set after the above --include
, --exclude
and --sort_by
operations are done. The 2 int values are applied on either side of the normal python slice syntax [int:int]
. These 2 values are expected to be integer values, but also can be the special string __NONE__
to specify a blank option in the slice.
These parameter examples show how they are used to generate the python Slice operation.
- Strip off the 1st 3 items in the list ==
[3:]
--slice 3 __NONE__
- Strip off the tail 4 items in the list ==
[:-4]
--slice __NONE__ -4
- Keep only the last 2 items in the list ==
[-2:]
--slice -2 __NONE__
- Skip the first 5 and take the next 10 only ==
[5:15]
--slice 5 15
- Strip off the first and last items ==
[1:-1]
--slice 1 -1
When the operation selected is one of the list
operations, set this option to true to generate a summary json output instead of the list of results output. This option is defaulted to true
when any of the delete
operations are selected.
- listPackages
default=false
- listPackageVersions
default=false
- deletePackageVersions
default=true
--summary true
Set to true
to generate a considerable amount of additional debugging log outputs. Helps when there are unknown reasons for unexpected results
--debug true
This option only affects any of the destructive operations. true
by default, this allows you to test any of the delete
operations expected results. Set this to false
to actually execute the delete
operation. Be careful when enabling this option.
--dryrun false
Limit the before-filtering initial fetching total from the github api to this maximum total amount. The GitHub api uses a paging mechanism to fetch all data. The github imposed limit is 100 records at a time per page. To prevent an accidental over-load of fetching, this utility has a default total initial fetch limit of 1000
records. You can artificially limit this to a shorter amount for batching or testing purposes. Or, you can increase it, at your own risk.
Important not here, this limit is imposed BEFORE any filtering or sorting is done. This simply limits the data loaded that is then to be processed by the filters and sorting of this utility.
--fetch_limit 10
--fetch_limit 50
When this action runs, the various options run in a particular order. Allowing for predictable results.
- fetch all records in default order from GitHub. Up to the default maximum 1000
--fetch_limit
- stop fetch once the optional
--fetch_limit
is reached. - apply the
--include
filter. - apply the
--exclude
filter. - apply the
--sort
and--reverse
operation. - excute the operation on the final list
A standard string
value to be passed in. No need to wrap in ""
quotes.
A standard int
value.
A string value, one of true
or false
.
This is a string that represents the json-path
to a value in the data results. This action uses the jsonpath-ng
library to reference a value in the json api results.
a standard python regex
string. Online regex testers help a great deal in designing a good regex
for your needs. This action uses the standard python re
package
A special values that signifies a blank/empty/null/None value
The power of this action is in it's ability to filter by any value within the api response that suites your needs. You are also not limited to --include
, but can also use --exclude
at the same time.
Both the --include
and --exclude
options take 2 parameters. With the 1st being a json-path
to a data field in the response list. The second parameter is a regex
string used to match with the content at the json-path
provided.
Rather than have this action define a set of filter criteria for you to chose from, instead I went with allowing you to define the filter to fit your needs. With this great power, comes great responsibility!
First you will need to understand the response format of the json returned by the --operation
. This json
response is needed to build a json-path
to fit your filter needs. See sample json responses, or run one of the list --operations
with --fetch_limit 10
to see what fields are in the response json.
Once you know the field you wish to filter by, you must now define a regex
string to match the field value to suite your filter needs.
Finally, you must decide if the result you want will --include
or --exclude
objects that match the filter parameters. Or, a combination of both together.
The response json from the api come in one of 2 forms. Either a single object, or a list of objects. The listPacakges
and listPackageVersions
operations both return lists of objects. The json path to this list automaticaly includes the leading [*]
of the json-path
The json-path
format generally takes the form of a dot-notation between model objects within the json.
{ "id": 1, "name": "foo", "sub": { "subid": "A", "name": "bar" } }
To reference the "bar"
value at the sub models "name"
field, you need simply sub.name
Explaining Regex matching here is beyond the scope of this readme. You must test various regex strings to be sure you get teh response you expect, without making false positive matches for things you mean to avoid.
Container tags are kept within a sub-model of each versions json model. The container tags are contained within an array/list of strings. You must include the trailing [*]
to have the filters match all tags. if you leave the trailing [*]
off, the regex will attempt to match against the resulting list, which will not work as expected.
metadata.container.tags[*]
== All tags (this is the normal scenario)metadata.container.tags[0]
== Only the 1st tag (I would not use this, no guarantee of the order the tags are stored in the list)
Both package and package version models include a root model level created_at
timestamp in an ISO8601 format.
created_at
Like the created timestamp, both package and version models include a root model level updated_at
timestamp in ISO8601 format.
updated_at
Both packages and package versions have a unique ID at the root model.
id
Provided here is a set of somewhat common package administration tasks. Each task listed provides a command line (CLI) version, the Action version, and a local Docker run version.
You can also reference this task in action here in this repo.
Use a --include
filter matching your tag. Include the --summary
flag to create a simple json output. For example, if you want to report on all package versions with -latest
at the end of any tags. Such as tags with develop-latest
and v1-latest
.
pipenv run ghpkadmin --operation listPackageVersions --ghtoken <token> --org <your org> --package_type container --package_name <name> --include "metadata.container.tags[*]" ".+-latest" --summary
- name: List Packages with tag
uses: shaneapowell/ghaction-package-admin@v1
with:
operation: listPackageVersions
ghtoken: ${{ github.token }}
org: ${{ github.repository_owner }}
page_type: container
package_name: <name>
include: metadata.container.tags[*]" .+-latest$
summary: true
docker run --name ghpkgadmin --rm ghcr.io/shaneapowell/ghaction-package-admin:v1 --operation listPackageVersions --ghtoken <token> --org <your org> --package_type container --package_name <name> --include "metadata.container.tags[*]" ".*-latest" --summary
Use the --exclude
filter matching any that include any value in the tags. In other words, only return versions that do NOT have a value in the tags list.
"metadata.container.tags[*]"
-> Finds the values of all tags".+"
-> Matches to any content within any tag. You could also use".*"
since the tags are held within a list of strings.
pipenv run ghpkadmin --operation deletePackageVersions --ghtoken <token> --org <your org> --package_type container --package_name <name> --exclude "metadata.container.tags[*]" ".+"
- name: Delete Packages without Tags
uses: shaneapowell/ghaction-package-admin@v1
with:
operation: deletePackageVersions
ghtoken: ${{ github.token }}
org: ${{ github.repository_owner }}
page_type: container
package_name: <name>
exclude: metadata.container.tags[*] .+
docker container run --name ghpkgadmin --rm ghcr.io/shaneapowell/ghaction-package-admin:v1 --operation deletePackageVersions --ghtoken <token> --org <your org> --package_type container --package_name <name> --exclude "metadata.container.tags[*]" ".*"
Lets include only versions with DEVELOP-
. And lets also make sure to avoid any versions with -latest
anywhere in the tag.
- Use the
--include
filter to find only the versions that havedevelop-
- Lets use the regex
(?i)
to make the filter case-insensitive. Matchingdevelop
andDevelop
andDEVELOP
- Use the
--exclude
filter to avoid tags we find with-latest
in the text. - Sort the result by either
created_at
orupdated_at
timestamps in reverse order youngest at the top. - Slice the result to skip the first 5.
pipenv run ghpkadmin --operation deletePackageVersions --ghtoken <token> --org <your org> --package_type container --package_name <name> --include "metadata.container.tags[*]" "(?i)DEVELOP-.*" --exclude "metadata.container.tags[*] .*-latest --sort updated_at --reverse --slice 5 __NONE__
- name: Delete all but the most recent packages for a tag (case insensitive).
uses: shaneapowell/ghaction-package-admin@v1
with:
operation: deletePackageVersions
ghtoken: ${{ github.token }}
user: ${{ github.repository_owner }}
page_type: container
package_name: <name>
include: metadata.container.tags[*]" (?i)^DEVELOP-.*
exclude: metadata.container.tags[*]" .*-latest$
sort: "updated_at"
reverse: true
slice: 5 __NONE__
The above common tasks examples include the action
syntax, a cli
syntax, and a docker run
syntax. You can reference the developer doc for some more help in how to try the cli
or docker
examples on your local machine.
The delete operation is done 1 record at a time. Each operation itself takes a few seconds to complete. If you are trying to clean out hundreds of records at once, this WILL take quite some time to complete. You might want to run in batches by specifying the --slice 20 __NONE__
option to only delete 20 at a time. Or the --fetch_limit
option. You might need to run your delete operation locally in batches to widdle down your list.
See the list of sample json responses for reference.
Or, run one of the --operation
commands to list the contents, using the --fetch_limit 10
to simply return the first 10 records and stop.
By default the --dryrun
flag is set to True
. You MUST explicitly turn it off with
--dryrun false
Typically caused by your PAT
token not having the necessary authorization permissions to perform the operation.
99% of the time, this is a permissions issue. Ensure the user account tied to your github token, has admin
rights on the package in question.
This could very well be a bug in this action. If you receive one of these errors, please submit an issue
with debugging output enabled to assist with resolving the problem.