Access additional license data #23
Conversation
SPDX version upgraded to the 3.17 version. After the upgrade the licensedb.Detect() function begun to panic with: panic: runtime error: index out of range [0] with length 0. The root cause was in the queryLicenseAbstract method of the database. After adding a length check to amount of matches found the issue disappeared. The upgrade also affected a confidence of the matched licenses in the TestCmdMain, so they were adjusted accordingly to pass.
In order to extend functionality the two new exportable functions were added to get more info about the license. LicenseURLs returns a list of URLs for the license and LicenseName returns a full name of the license. Both accepts a license identifier as an input parameter and do search over the license database. If there is no info about provided license the error ErrUnknownLicenseID is returned. The extract_urls.go was extended to produce one more column in urls.csv file that includes schema of the URL. It is needed to construct the full URL that will be returned from the LicenseURLs.
|
@bzz could you please take a look at this PR? |
|
Hi @lafriks! Thanks for taking an eye on my changes. Could you please explain why |
|
Seems like Lint job fails because Update: I have increased version of the golangci-lint tool to 1.47.0 version and fixed the issues. |
|
I don't think I have ever supported windows, it's ok to remove the CI checks for that OS. |
The Race action executes tests with -race flag and sometimes can take more than 10m which is the default test timeout. To mitigate that possibility the timeout is increased to 60m by explicit usage of the -timeout flag.
The version of the golangci-lint tool upgraded to a newer 1.47.0 version. As there were couple of lint issues (unsafeptr: possible misuse of reflect.SliceHeader (govet)), those were fixed to pass the static code check.
Widnows was not ever explicitly supported by the author and causes troubles on CI action runs. Those we drop it as redundant.
pavelmemory
force-pushed
the
access-additional-license-data
branch
from
8fedebb
to
a7c9e28
Compare
@vmarkovtsev I have removed Windows OS from the actions, hope the CI will pass this time. |
|
Hey @vmarkovtsev, is there anything we can do to help get this PR over the finish line? @pavelmemory made the necessary changes to drop CI for Windows (this would actually close #22). |
|
As CI is green now and there are new tests and review ✅ - I'll be merging this, if there is no further discussion. As this change includes new API - it's a bit hard for me to say off hand how backward-compatible is it, and if it is - I assume v4.4.0 needs to be released, otherwise, it's v5.0.0. WDYT? |
|
API seems to be backward compatible as only adds new methods so imho v4.4.0 is way to go |
|
Closed&re-opened to trigger the CI, going to merge as soon as it passes. |
|
Linter complains on the |
|
...wonder what makes Race CI profile run 20min here, while it's only 6m in #24 🤔 |
The change in this PR addresses a problem described in the issue "Make license database public".
In order to get access to additional license data the two new exportable functions were added:
The values are get from the JSON representation of the license stored in the https://github.com/spdx/license-list-data/archive/v$(SPDX_DATA_VERSION).tar.gz as all other information about the licenses. But to get the fully qualified URLs the third column was added to the urls.csv file and it contains schema of the URL. That is done to have possibility re-build the full URL and do not change existing format of the second column that may be used by someone already.
Both functions return a new
ErrUnknownLicenseIDerror in case the license ID provided as a parameter doesn't exist in the database.The change also includes upgrade of the SPDX_DATA_VERSION to the latest at the moment 3.17 version. After the update the tool begun to panic because of the index access to the empty slice. The fix has been applied to solve that issue.
After the upgrade the expected confidence in the tests were adjusted to the new results.
The one question that remains open here is why
.mit-license.orgliteral is used in thelicensedb/internal/assets/extract_urls.go? Should it be changed to the://mit-license.orgto have a proper URL format?