Add a tags.json meta item, require it to be list of strings

kart/base_dataset.py

@@ -403,11 +403,9 @@ def _apply_meta_delta_to_tree(
 
         # Actual implementation once we've figured out there's no conflict:
         if new_value is not None:
-            if name.endswith(".json"):
-                new_value = json_pack(new_value)
-            else:
-                new_value = ensure_bytes(new_value)
-            object_builder.insert(name, new_value)
+            definition = self.get_meta_item_definition(name)
+            file_type = MetaItemFileType.get_from_definition_or_suffix(definition, name)
+            object_builder.insert(name, file_type.encode_to_bytes(delta.new_value))
         else:
             object_builder.remove(name)
 

kart/meta.py

@@ -181,6 +181,8 @@ def meta_set(ctx, message, dataset, items):
         message = f"Update metadata for {dataset}"
 
     def _parse(key, value):
+        if value == "":
+            return None
         value = value_optionally_from_text_file(value, key, ctx, encoding="utf-8")
         if key.endswith(".json"):
             try:

kart/meta_items.py

@@ -7,6 +7,31 @@
 from kart.serialise_util import ensure_text, ensure_bytes, json_pack, json_unpack
 
 
+class TagsJsonFileType:
+    # schema.json should be checked on read and write, by dropping any optional fields that are None.
+    def decode_from_bytes(self, data):
+        if data is None:
+            return None
+        return self.assert_list_of_strings(json_unpack(data))
+
+    def encode_to_bytes(self, meta_item):
+        if meta_item is None:
+            return None
+        return json_pack(self.assert_list_of_strings(meta_item))
+
+    def assert_list_of_strings(self, meta_item):
+        try:
+            assert isinstance(meta_item, list)
+            for tag in meta_item:
+                assert isinstance(tag, str)
+        except AssertionError as e:
+            raise AssertionError("tags.json should be a list of strings")
+        return meta_item
+
+
+TagsJsonFileType.INSTANCE = TagsJsonFileType()
+
+
 class SchemaJsonFileType:
     # schema.json should be normalised on read and write, by dropping any optional fields that are None.
     def decode_from_bytes(self, data):
@@ -156,6 +181,9 @@ def match_group(self, meta_item_path, match_group):
 # A longer description about the dataset's contents:
 DESCRIPTION = MetaItemDefinition("description", MetaItemFileType.TEXT)
 
+# A list of tags - each tag is free form text.
+TAGS_JSON = MetaItemDefinition("tags.json", TagsJsonFileType.INSTANCE)
+
 # JSON representation of the dataset's schema. See kart/tabular/schema.py, datasets_v3.rst
 SCHEMA_JSON = MetaItemDefinition("schema.json", SchemaJsonFileType.INSTANCE)
 

kart/point_cloud/v1.py

@@ -52,6 +52,7 @@ class PointCloudV1(BaseDataset):
 
     TITLE = meta_items.TITLE
     DESCRIPTION = meta_items.DESCRIPTION
+    TAGS_JSON = meta_items.TAGS_JSON
     METADATA_XML = meta_items.METADATA_XML
 
     # Which tile format(s) this dataset requires / allows.
@@ -63,6 +64,7 @@ class PointCloudV1(BaseDataset):
     META_ITEMS = (
         TITLE,
         DESCRIPTION,
+        TAGS_JSON,
         METADATA_XML,
         FORMAT_JSON,
         SCHEMA_JSON,

kart/tabular/v3.py

@@ -69,6 +69,7 @@ class TableV3(RichTableDataset):
     # === Visible meta-items ===
     TITLE = meta_items.TITLE
     DESCRIPTION = meta_items.DESCRIPTION
+    TAGS_JSON = meta_items.TAGS_JSON
     METADATA_XML = meta_items.METADATA_XML
     SCHEMA_JSON = meta_items.SCHEMA_JSON
     CRS_DEFINITIONS = meta_items.CRS_DEFINITIONS
@@ -92,6 +93,7 @@ class TableV3(RichTableDataset):
     META_ITEMS = (
         TITLE,
         DESCRIPTION,
+        TAGS_JSON,
         METADATA_XML,
         SCHEMA_JSON,
         CRS_DEFINITIONS,

kart/tabular/working_copy/base.py

@@ -15,6 +15,7 @@
     NotYetImplemented,
 )
 from kart.key_filters import DatasetKeyFilter, FeatureKeyFilter, RepoKeyFilter
+from kart import meta_items
 from kart.promisor_utils import LibgitSubcode
 from kart.sqlalchemy.upsert import Upsert as upsert
 from kart.tabular.table_dataset import TableDataset
@@ -42,6 +43,14 @@ class TableWorkingCopy(WorkingCopyPart):
     self.kart_tables - sqlalchemy Table definitions for kart_state and kart_track tables.
     """
 
+    # Subclasses should override if they can support more meta-items eg description or metadata.xml
+    SUPPORTED_META_ITEMS = (
+        meta_items.TITLE,
+        meta_items.SCHEMA_JSON,
+        meta_items.CRS_DEFINITIONS
+        # Not description, not metadata.xml, except where overridden by a subclass.
+    )
+
     @property
     def WORKING_COPY_TYPE_NAME(self):
         """Human readable name of this type of working copy, eg "PostGIS"."""
@@ -505,6 +514,16 @@ def _remove_hidden_numeric_diffs(
             if old_col_dict.get(name) is None and new_col_dict.get(name) == default_val:
                 new_col_dict[name] = None
 
+    def _is_supported_meta_item(self, meta_item_name):
+        if meta_item_name in self.SUPPORTED_META_ITEMS:
+            return True
+        for definition in self.SUPPORTED_META_ITEMS:
+            if not isinstance(definition, meta_items.MetaItemDefinition):
+                continue
+            if definition.matches(meta_item_name):
+                return True
+        return False
+
     def _remove_hidden_meta_diffs(self, dataset, ds_meta_items, wc_meta_items):
         """
         Remove any meta diffs that can't or shouldn't be committed, and so shouldn't be shown to the user.
@@ -516,6 +535,11 @@ def _remove_hidden_meta_diffs(self, dataset, ds_meta_items, wc_meta_items):
         def _safe_del(dict_, key):
             dict_.pop(key, None)
 
+        ds_meta_keys = list(ds_meta_items.keys())
+        for key in ds_meta_keys:
+            if not self._is_supported_meta_item(key):
+                _safe_del(ds_meta_items, key)
+
         # A dataset should have at most ONE of "metadata.xml" or "metadata/dataset.json".
         # The XML file is newer and supercedes the JSON file.
         # The GPKG adapter generates both, so we delete one so as to match the dataset.

kart/tabular/working_copy/gpkg.py

@@ -11,6 +11,7 @@
 import sqlalchemy as sa
 from kart import crs_util
 from kart.exceptions import InvalidOperation
+from kart import meta_items
 from kart.sqlalchemy import text_with_inlined_params
 from kart.sqlalchemy.adapter.gpkg import KartAdapter_GPKG
 from kart.schema import Schema
@@ -34,6 +35,15 @@ class WorkingCopy_GPKG(TableWorkingCopy):
 
     WORKING_COPY_TYPE_NAME = "GPKG"
 
+    SUPPORTED_META_ITEMS = (
+        meta_items.TITLE,
+        meta_items.DESCRIPTION,
+        meta_items.SCHEMA_JSON,
+        meta_items.CRS_DEFINITIONS,
+        meta_items.METADATA_XML,
+        "metadata/dataset.json",
+    )
+
     def __init__(self, repo, location):
         self.repo = repo
         self.path = self.location = location
@@ -93,7 +103,7 @@ def normalise_location(cls, wc_location, repo):
 
     @property
     def full_path(self):
-        """ Return a full absolute path to the working copy """
+        """Return a full absolute path to the working copy"""
         return (self.repo.workdir_path / self.path).resolve()
 
     @property

kart/tabular/working_copy/mysql.py

@@ -228,20 +228,9 @@ def try_align_schema_col(cls, old_col_dict, new_col_dict):
 
         return new_type == old_type
 
-    _UNSUPPORTED_META_ITEMS = (
-        "description",
-        "metadata/dataset.json",
-        "metadata.xml",
-    )
-
     def _remove_hidden_meta_diffs(self, dataset, ds_meta_items, wc_meta_items):
         super()._remove_hidden_meta_diffs(dataset, ds_meta_items, wc_meta_items)
 
-        # Nowhere to put these in SQL Server WC
-        for key in self._UNSUPPORTED_META_ITEMS:
-            if key in ds_meta_items:
-                del ds_meta_items[key]
-
         for key in ds_meta_items.keys() & wc_meta_items.keys():
             if not key.startswith("crs/"):
                 continue

kart/tabular/working_copy/postgis.py

@@ -250,9 +250,6 @@ def _suspend_triggers(self, sess, dataset):
             """
         )
 
-    # Postgis has nowhere obvious to put this metadata.
-    _UNSUPPORTED_META_ITEMS = ("description", "metadata/dataset.json", "metadata.xml")
-
     # PostGIS approximates an int8 as an int16 - see super()._remove_hidden_meta_diffs
     @classmethod
     def try_align_schema_col(cls, old_col_dict, new_col_dict):
@@ -275,14 +272,6 @@ def try_align_schema_col(cls, old_col_dict, new_col_dict):
 
         return new_type == old_type
 
-    def _remove_hidden_meta_diffs(self, dataset, ds_meta_items, wc_meta_items):
-        super()._remove_hidden_meta_diffs(dataset, ds_meta_items, wc_meta_items)
-
-        # Nowhere to put these in postgis WC
-        for key in self._UNSUPPORTED_META_ITEMS:
-            if key in ds_meta_items:
-                del ds_meta_items[key]
-
     def _is_builtin_crs(self, crs):
         auth_name, auth_code = crs_util.parse_authority(crs)
         return auth_name in ("EPSG", "ESRI") or auth_code == "900913"  # GOOGLE

kart/tabular/working_copy/sqlserver.py

@@ -263,20 +263,9 @@ def try_align_schema_col(cls, old_col_dict, new_col_dict):
 
         return new_type == old_type
 
-    _UNSUPPORTED_META_ITEMS = (
-        "description",
-        "metadata/dataset.json",
-        "metadata.xml",
-    )
-
     def _remove_hidden_meta_diffs(self, dataset, ds_meta_items, wc_meta_items):
         super()._remove_hidden_meta_diffs(dataset, ds_meta_items, wc_meta_items)
 
-        # Nowhere to put these in SQL Server WC
-        for key in self._UNSUPPORTED_META_ITEMS:
-            if key in ds_meta_items:
-                del ds_meta_items[key]
-
         # Nowhere to put custom CRS in SQL Server, so remove custom CRS diffs.
         # The working copy doesn't know the true authority name, so refers to them all as CUSTOM.
         # Their original authority name could be anything.