Descarteslabs

Python Version Support
- Python 3.8 is now supported in the client.
- As Python 3.5 has reached End Of Life, it is no longer supported by the descarteslabs client.

Tasks Client
- Altered the behavior of Task function creation. Deprecation warnings will be issued when attempting to create
a Task function for which support will be removed in the near future. **It is
strongly recommended to test any existing code which uses the Tasks client when upgrading to this
release.**
- New tasks public images for for use with Python 3.8 are available.

Workflows (channel `v0-17`) - Fixed
- `.pick_bands` supports proxy `wf.Str` objects; `.unpack_bands` supports `wf.Str` and `wf.Tuple[wf.Str, ...]`.
- Better performance constructing a `wf.Array` from a `List` of numbers (like `wf.Array(ic.sum(["pixels", "bands"]))`)
- No more error using `wf.publish` as a decorator on a function without a docstring

Fixed
No more irrelevant `DeprecationWarning`s when importing the `descarteslabs` package ([235](https://github.com/descarteslabs/descarteslabs-python/issues/235)). Deprecated functionality in the package will now show `FutureWarning`s instead.

Workflows (channel `v0-16`) - Fixed
- `wf.map.geocontext` doesn't raise an error about the CRS of the map
- `wf.flows` doesn't raise an error about versions from incompatible channels

Catalog client

- Example code has been cleaned up.

Workflows (channel `v0-16`) - Added
- **Sharing of any Workflows object as a `Workflow`** with version and access control. Browse through shared `Workflow`s with the `wf.flows` browser widget.
- **Upload images to the DL catalog from Workflows jobs**. Usage details can be found [in the docs](https://docs.descarteslabs.com/descarteslabs/workflows/docs/destinations.html).
- `wf.np.median`
- `Job.cancel()` to cancel running jobs.
- Transient failures in Jobs are automatically retried, resulting in fewer errors.
- [Search widget](https://ipyleaflet.readthedocs.io/en/latest/api_reference/search_control.html) on `wf.map` by default.

Workflows - Fixed
- Bitwise operations on imagery no longer fail
- `wf.np.linspace` no longer fails when being called correctly
- `.median` is slightly less prone to OOM errors

Workflows - Changed
- `Array.to_imagery` now accepts `KnownDict` for bandinfo and properties.
- `Number`s can now be constructed from `Str`s

Workflows (channel `v0-15`) - Added
- **Output formats for `.compute` including GeoTIFF, JSON**, PyArrow, and MessagePack. Usage details can be found [in the docs](https://docs.descarteslabs.com/descarteslabs/workflows/docs/formats.html).
- **Destinations for Job results: download and email**. Usage details can be found [in the docs](https://docs.descarteslabs.com/descarteslabs/workflows/docs/destinations.html).
- **Save `.compute` outputs to a file** with the `file=` argument.
- **Pixel value inspector**: click in the map widget to view pixel values.
- **`wf.ifelse`** for simple conditional logic.
- NumPy functions including `hypot`, `bitwise_and`, `bitwise_or`, `bitwise_xor`, `bitwise_not`, `invert`, and `ldexp`
- Bitwise `Array` and `MaskedArray` operations
- `size` attribute on `Array` and `MaskedArray`
- `astype` function on `Array` and `MaskedArray` for changing the dtype
- `flatten` function on `Array` and `MaskedArray` for flattening into a 1D array
- `MaskedArray.compressed` for getting all unmasked data as a 1D array
- `get` function on `Dict` and `KnownDict` for providing a default value if a key does not exist
- `nbands` attribute on `Image` and `ImageCollection`
- `proxify` can handle `scenes.GeoContext`s
- `Dict.contains`, `Dict.length`

Workflows - Fixed
- **Fewer failures and hanging calls when connecting to the Workflows backend** (like `.compute`, `.visualize`, `Job.get`, etc.)
- **`wf.numpy.histogram` works correctly with computed values for `range` and `bins` (such as `range=[arr.min(), arr.max()]`)**
- More consistent throughput when a large number of jobs are submitted
- `Array`s can now be constructed from proxy `List`s
- `MaskedArray.filled` works correctly when passed Python values
- Long-running sessions (like Jupyter kernels) refresh credentials instead of failing with auth errors after many hours of use
- `wf.numpy.dot` and `wf.numpy.einsum` no longer fail when being called correctly
- Occasional errors like `('array-89199362e9a5d598fb5c82805136834d', 0, 0)` when calling `wf.compute()` with multiple values are resolved

Workflows - Changed
- **`pick_bands` accepts duplicate band names.** Enjoy easier Sentinel-1 `"vv vh vv"` visualizations!
- **`ImageCollection.from_id` is always ordered by date**
- `wf.numpy.percentile` no longer accepts an `axis` argument
- **breaking** `wf.Job` construction and interface changes:
- Use a single `wf.Job(..)` call instead of `wf.Job.build(...).execute()` to create and launch a Job
- New `Job.result_to_file` method
- `Job.status` is removed in favor of a single `Job.stage`
- `wf.TimeoutError` renamed to `wf.JobTimeoutError`

Workflows (channel `v0-14`) - Added
- **191 functions from NumPy are available for Workflows `Array`s**, including parts of the `numpy.linalg` and `numpy.ma` submodules. See the full list [on the docs](https://docs.descarteslabs.com/descarteslabs/workflows/docs/types/numpy.html).
- `index_to_coords` and `coords_to_index` methods on `Image`/`ImageCollection`/`GeoContext` for converting between geospatial and array coordinates
- `value_at` function on `Image` and `ImageCollection` for extracting single pixel values at spatial coordinates.

Workflows - Fixed
- Using datetimes as parameters to `visualize` behaves correctly.

Catalog client

- Fixed a bug that prevented uploading ndarrays of type `uint8`

Workflows (channel `v0-13`) - Added
- Array support for `argmin`, `argmax`, `any`, `all`
- `pick_bands` supports an `allow_missing` kwarg to drop band names that may be missing from the data without an error.
- `wf.compute` supports passing lists or tuples of items to compute at the same time. Passing multiple items to `wf.compute`, rather than calling `obj.compute` for each separately, is usually faster.
- Casting from `Bool` to `Int`: `wf.Int(True)`
- Experimental `.inspect()` method for small computations during interactive use.

Workflows - Changed
- **[breaking]** Array no longer uses type parameters: now you construct an Array with `wf.Array([1, 2, 3])`, not `wf.Array[wf.Int, 1]([1, 2, 3])`. Remember, Array is an experimental API and will continue to make frequent breaking changes!
- Workflows now reuses the same gRPC client by default---so repeated or parallel calls to `.compute`, etc. will be faster. Calling `.compute` within a thread pool will also be significantly more efficient.

Workflows - Fixed
- `wf.numpy.histogram` correctly accepts a `List[Float]` as the `range` argument