changelog and migration guide entries

CHANGELOG.md

@@ -27,6 +27,7 @@ methods are `json` and `yaml` serializable.
 - `CogniteResource.to_pandas` and `CogniteResourceList.to_pandas` now converts known timestamps to `datetime` by
   default. Can be turned off with the new parameter `convert_timestamps`. Note: To comply with older pandas v1, the
   dtype will always be `datetime64[ns]`, although in v2 this could have been `datetime64[ms]`.
+- `CogniteImportError` can now be caught as `ImportError`.
 
 ### Deprecated
 - The Templates API (migrate to Data Modeling).
@@ -38,6 +39,10 @@ with no easy way to add a prefix. Also, it no longer expands metadata by default
 - Additionally, `Asset.to_pandas`, now accepts the parameters `expand_aggregates` and `aggregates_prefix`. Since
   the possible `aggregates` keys are known, `camel_case` will also apply to these (if expanded) as opposed to
   the metadata keys.
+- More narrow exception types like `CogniteNotFoundError` and `CogniteDuplicatedError` are now raised instead of
+  `CogniteAPIError` for the following methods: `DatapointsAPI.retrieve_latest`, `RawRowsAPI.list`,
+  `RelationshipsAPI.list`, `SequencesDataAPI.retrieve`, `SyntheticDatapointsAPI.query`. Additionally, all calls
+  using `partitions` to API methods like `list` (or the generator version) now do the same.
 - The `CogniteResource._load` has been made public, i.e., it is now `CogniteResource.load`.
 - The `CogniteResourceList._load` has been made public, i.e., it is now `CogniteResourceList.load`.
 - All `.delete` and `.retrieve_multiple` methods now accepts an empty sequence, and will return an empty `CogniteResourceList`.

MIGRATION_GUIDE.md

@@ -50,6 +50,10 @@ Changes are grouped as follows:
 ### Changed
 - All `assert`s meant for the SDK user, now raise appropriate errors instead (`ValueError`, `RuntimeError`...).
 - `CogniteAssetHierarchyError` is no longer possible to catch as an `AssertionError`.
+- More narrow exception types like `CogniteNotFoundError` and `CogniteDuplicatedError` are now raised instead of
+  `CogniteAPIError` for the following methods: `DatapointsAPI.retrieve_latest`, `RawRowsAPI.list`,
+  `RelationshipsAPI.list`, `SequencesDataAPI.retrieve`, `SyntheticDatapointsAPI.query`. Additionally, all calls
+  using `partitions` to API methods like `list` (or the generator version) now do the same.
 - Several methods in the data modelling APIs have had parameter names now correctly reflect whether they accept
   a single or multiple items (i.e. id -> ids).
 - Loading `ObjectDetection` attributes `.attributes`, `.bounding_box`, `.polygon` and
@@ -83,6 +87,9 @@ Changes are grouped as follows:
 - The attribute `columns` in `Sequence` has been changed from `SequenceType[dict]` to `SequnceColumnList`.
 -The class `SequenceRows` in `client.data_classes.transformations.common` has been renamed to `SequenceRowsDestination`.
 
+### Optional
+- `CogniteImportError` can now be caught as `ImportError`.
+
 ## From v5 to v6
 ### Removed
 - Removed support for legacy auth (API keys, service accounts, client.login.status()). Use OIDC to authenticate instead and use `client.iam.token.inspect()` instead of `login.status()`.