Internetarchive

++++++++++++++++++

**Bugfixes**

- Fixed bug in search where search requests were failing with ``KeyError: 'items'``.

++++++++++++++++++

**Features and Improvements**

- Added support for adding custom headers to ``ia search``.

**Bugfixes**

- ``internetarchive.utils.get_s3_xml_text()`` is used to parse errors returned by S3 in XML.
Sometimes there is no XML in the response.
Most of the time this is due to 5xx errors.
Either way, we want to always return the HTTPError, even if the XML parsing fails.
- Fixed a regression where ``:`` was being stripped from filenames in upload.
- Do not create a directory in ``download()`` when ``return_responses`` is ``True``.
- Fixed bug in upload where file-like objects were failing with a TypeError exception.

++++++++++++++++++

**Bugfixes**

- Fixed bug in ``ia upload`` where all commands would fail if multiple collections were specified (e.g. -m collection:foo -m collection:bar).

++++++++++++++++++

**Features and Improvements**

- Loosened up ``jsonpatch`` requirements, as the metadata API now supports more recent versions of the JSON Patch standard.
- Added support for building "snap" packages (https://snapcraft.io/).

**Bugfixes**

- Fixed bug in upload where users were unable to add their own timeout via ``request_kwargs``.
- Fixed bug where files with non-ascii filenames failed to upload on some platforms.
- Fixed bug in upload where metadata keys with an index (e.g. ``subject[0]``) would make the request fail if the key was the only indexed key provided.
- Added a default timeout to ``ArchiveSession.s3_is_overloaded()``.
If it times out now, it returns ``True`` (as in, yes, S3 is overloaded).

++++++++++++++++++

**Features and Improvements**

- Added 60 second timeout to all upload requests.
- Added support for uploading empty files.
- Refactored ``Item.get_files()`` to be faster, especially for items with many files.
- Updated search to use IA-S3 keys for auth instead of cookies.

**Bugfixes**

- Fixed bug in upload where derives weren't being queued in some cases where checksum=True was set.
- Fixed bug where ``ia tasks`` and other ``Catalog`` functions were always using HTTP even when it should have been HTTPS.
- ``ia metadata`` was exiting with a non-zero status for "no changes to xml" errors.
This now exits with 0, as nearly every time this happens it should not be considered an "error".
- Added unicode support to ``ia upload --spreadsheet`` and ``ia metadata --spreadsheet`` using the ``backports.csv`` module.
- Fixed bug in ``ia upload --spreadsheet`` where some metadata was accidentally being copied from previous rows
(e.g. when multiple subjects were used).
- Submitter wasn't being added to ``ia tasks --json`` output, it now is.
- ``row_type`` in ``ia tasks --json`` was returning integer for row-type rather than name (e.g. 'red').

++++++++++++++++++

**Features and Improvements**

- Added option to download() for returning a list of response objects
rather than writing files to disk.