fix bug in …/groupBy/tag and …/groupBy/key endpoints when key/values do not exist in keytables

[tyrasd]

Fixes a bug/crash in …/groupBy/tag requests triggered when using non-existing groupByKey or groupByValues are used in a query.

To Reproduce

curl -X GET "https://api.ohsome.org/v1/elements/count/groupBy/tag?bboxes=8.67%2C49.39%2C8.71%2C49.42&filter=type%3Away%20and%20building%3D*&format=json&groupByKey=building&groupByValues=DoesNotExist1&time=2023-01-01" -H  "accept: application/json"

This query is requesting results grouped by the tag building and value DoesNotExist1 (which does not exits in the keytables). The result, however contains building=yes as the groupByObject. When performing a query with two non-existing groupByValues (e.g. DoesNotExist1,DoesNotExist2), the results includes two entries with remainder as groupByObject. When requesting three or more non-existing groupByValues, the query crashes and a 500 server error is returned.

Similar bugs occur when performing a query with a non-existing tag key as the groupByKey.

The Problem and Fix

The issue is caused by a mistake in the code, where consecutive negative tag numbers are used as keyInt/valuesInt for OSM tags which do not exist in the keytables, but the hardcoded value -1 is also used to represent the remainder portion of the result. This is causing the mixup between the value DoesNotExist1 and the superfluous remainder entry in the example result above. Further, the negative pseudo-indices started at 0 which is also the index of the most common tag value for the corresponding tag. This is causing the mixup between the value DoesNotExist2 and the result entry for the value yes in the example above. The crashes for any additional non-existing values are caused by a null pointer exception where the (still) non-existing pseudo-indices are attempted to be converted back to strings using the tag translator.

This is fixed by starting the negative pseudo-indices at position -2 where confusion with the -1 value for the remainder entry is possible and using the input values to resolve the pseudo-indices back to strings.

Checklist

• My code follows the code-style rules, and I have checked on the static analyses
• I have commented my code
• I have written javadoc (required for public methods)
• I have added sufficient unit and API tests
• I have made corresponding changes to the documentation
• I have updated the CHANGELOG.md
• I have adjusted the examples in the check-ohsome-api script or created an issue in the corresponding repository. More Information here.

[rtroilo]

Thank you