exchangelib

• Fix access to shared folders

• Remove timezone warnings in GetUserAvailability
• Update NoVerifyHTTPAdapter for newer requests versions

• Fix traversal of public folders in Account.public_folders_root
• Mark certain distinguished folders as only supported on newer Exchange versions
• Fetch all autodiscover information by default

• Add O365InteractiveConfiguration helper class to set up MSAL auth for O365.
• Add exchangelib[msal] installation flavor to match the above.
• Various bug fixes related to distinguished folders.

• Fix various issues related to public folders and archive folders
• Support read-write for Contact.im_addresses
• Improve reporting of inbox rule validation errors

• Fix ErrorAccessDenied: Not allowed to access Non IPM folder caused by recent changes in O365.
• Add more intuitive API for inbox rules
• Fix various bugs with inbox creation

• Allow setting a custom Configuration.max_conections in autodiscover mode
• Add support for inbox rules. See documentation for examples.
• Fix shared folder access in delegate mode
• Support subscribing to all folders instead of specific folders

• Fix QuerySet operations on shared folders
• Fix globbing on patterns with more than two folder levels
• Fix case sensitivity of "/" folder navigation
• Multiple improvements related to consistency and graceful error handling

• Bugfix release

• Fix bug where certain folders were being assigned the wrong Python class.

• Fix PyPI package. No source code changes.

• Make SOAP-based autodiscovery the default, and remove support for POX-based discovery. This also removes support for autodiscovery on Exchange 2007. Only Account(..., autodiscover=True) is supported again.
• Deprecated RetryPolicy.may_retry_on_error. Instead, add custom retry logic in RetryPolicy.raise_response_errors.
• Moved exchangelib.util.RETRY_WAIT to BaseProtocol.RETRY_WAIT.

• Added support for SOAP-based autodiscovery, in addition to the existing POX (plain old XML) implementation. You can specify the autodiscover implementation explicitly using the autodiscover argument: Account(..., autodiscover="soap") or Account(..., autodiscover="pox"). POX is still the default.

• Added new OAuth2LegacyCredentials class to support username/password auth over OAuth.

• Fixed token refresh bug with OAuth2 authentication, again

• Fixed Protocol.get_free_busy_info() when called with +100 accounts.
• Allowed configuring DNS timeout for a single nameserver (Autodiscovery.DNS_RESOLVER_ATTRS["timeout""]) and the total query lifetime (Autodiscovery.DNS_RESOLVER_LIFETIME) separately.
• Fixed token refresh bug with OAuth2 authentication

• Bugfix release

• Bugfix release

• Fixed field name to match API: BaseReplyItem.received_by_representing to BaseReplyItem.received_representing
• Added fields received_by and received_representing to MeetingRequest, MeetingMessage and MeetingCancellation
• Fixed AppointmentStateField.CANCELLED enum value.

• Fixed issue where creating an Account with autodiscover and no config would never set a default retry policy.

• Fixed some spelling mistakes:
  • ALL_OCCURRENCIES to ALL_OCCURRENCES in exchangelib.items.base
  • Persona.orgnaization_main_phones to Persona.organization_main_phones
• Removed deprecated methods EWSTimeZone.localize(), EWSTimeZone.normalize(), EWSTimeZone.timezone() and QuerySet.iterator().
• Disambiguated chunk_size and page_size in querysets and services. Add a new QuerySet.chunk_size attribute and let it replace the task that QuerySet.page_size previously had. Chunk size is the number of items we send in e.g. a GetItem call, while page_size is the number of items we request per page in services like FindItem that support paging.
• Support creating a proper response when getting a notification request on the callback URL of a push subscription.
• FolderCollection.subscribe_to_[pull|push|streaming]() now return a single subscription instead of a 1-element generator.
• FolderCollection now has the same [pull|push|streaming]_subscription() context managers as folders.

• Fix filtering on array-type extended properties.
• Exceptions in GetStreamingEvents responses are now raised.
• Support affinity cookies for pull and streaming subscriptions.

• Support tzlocal>=4.1
• Bug fixes for paging in multi-folder requests.

• Support microsecond precision in EWSDateTime.ewsformat()
• Remove usage of the multiprocessing module to allow running in AWS Lambda
• Support tzlocal>=4

• Make FileAttachment.fp a proper BytesIO implementation
• Add missing CalendarItem.recurrence_id field
• Add SingleFolderQuerySet.resolve() to aid accessing a folder shared by a different account:

from exchangelib import Account
from exchangelib.folders import Calendar, SingleFolderQuerySet
from exchangelib.properties import DistinguishedFolderId, Mailbox

account = Account(primary_smtp_address="some_user@example.com", ...)
shared_calendar = SingleFolderQuerySet(
    account=account,
    folder=DistinguishedFolderId(
        id=Calendar.DISTINGUISHED_FOLDER_ID,
        mailbox=Mailbox(email_address="other_user@example.com"),
    ),
).resolve()

• Minor bugfixes

• Support updating items in Account.upload(). Previously, only insert was supported.
• Fixed types for Contact.manager_mailbox and Contact.direct_reports.
• Support getting text_body field on item attachments.

• Fixed bug when updating indexed fields on Contact items.
• Fixed bug preventing parsing of CalendarPermission items in the permission_set field.
• Add support for parsing push notification POST requests sent from the Exchange server to the callback URL.

• Add Folder.move() to move folders to a different parent folder.

• Add context managers Folder.pull_subscription(), Folder.push_subscription() and Folder.streaming_subscription() that handle unsubscriptions automatically.

• Move util._may_retry_on_error and and util._raise_response_errors to RetryPolicy.may_retry_on_error and RetryPolicy.raise_response_errors, respectively. This allows for easier customization of the retry logic.

• Add support for synchronization, subscriptions and notifications. Both pull, push and streaming notifications are supported. See https://ecederstrand.github.io/exchangelib/#synchronization-subscriptions-and-notifications

• Add a new max_connections option for the Configuration class, to increase the session pool size on a per-server, per-credentials basis. Useful when exchangelib is used with threads, where one may wish to increase the number of concurrent connections to the server.
• Add Message.mark_as_junk() and complementary QuerySet.mark_as_junk() methods to mark or un-mark messages as junk email, and optionally move them to the junk folder.
• Add support for Master Category Lists, also known as User Configurations. These are custom values that can be assigned to folders. Available via Folder.get_user_configuration().
• Persona objects as returned by QuerySet.people() now support almost all documented fields.
• Improved QuerySet.people() to call the GetPersona service if at least one field is requested that is not supported by the FindPeople service.
• Removed the internal caching in QuerySet. It's not necessary in most use cases for exchangelib, and the memory overhead and complexity is not worth the extra effort. This means that .iterator() is now a no-op and marked as deprecated. ATTENTION: If you previously relied on caching of results in QuerySet, you need to do you own caching now.
• Allow plain date, datetime and zoneinfo.ZoneInfo objects as values for fields and methods. This lowers the barrier for using the library. We still use EWSDate, EWSDateTime and EWSTimeZone for all values returned from the server, but these classes are subclasses of date, datetime and zoneinfo.ZoneInfo objects and instances will behave just like instance of their parent class.

• Change Kerberos dependency from requests_kerberos to requests_gssapi
• Let EWSDateTime.from_datetime() accept datetime.datetime objects with tzinfo objects that are dateutil , zoneinfo and pytz instances, in addition to EWSTimeZone.

• Allow overriding dns.resolver.Resolver class attributes via Autodiscovery.DNS_RESOLVER_ATTRS.

• Switch EWSTimeZone to be implemented on top of the new zoneinfo module in Python 3.9 instead of pytz . backports.zoneinfo is used for earlier versions of Python. This means that the ÈWSTimeZone methods timezone(), normalize() and localize() methods are now deprecated.
• Add EWSTimeZone.from_dateutil() to support converting dateutil.tz timezones to EWSTimeZone.
• Dropped support for Python 3.5 which is EOL per September 2020.
• Added support for CalendaItem.appointment_state, CalendaItem.conflicting_meetings and CalendarItem.adjacent_meetings fields.
• Added support for the Message.reminder_message_data field.
• Added support for Contact.manager_mailbox, Contact.direct_reports and Contact.complete_name fields.
• Added support for Item.response_objects field.
• Changed Task.due_date and Tas.start_date fields from datetime to date fields, since the time was being truncated anyway by the server.
• Added support for Task.recurrence field.
• Added read-only support for Contact.user_smime_certificate and Contact.ms_exchange_certificate. This means that all fields on all item types are now supported.

• Fix bug leading to an exception in CalendarItem.cancel().
• Improve stability of .order_by() in edge cases where sorting must be done client-side.
• Allow increasing the session pool-size dynamically.
• Change semantics of .filter(foo__in=[]) to return an empty result. This was previously undefined behavior. Now we adopt the behaviour of Django in this case. This is still undefined behavior for list-type fields.
• Moved documentation to GitHub Pages and auto-documentation generated by pdoc3.

• Remove use of ThreadPool objects. Threads were used to implement async HTTP requests, but were creating massive memory leaks. Async requests should be reimplemented using a real async HTTP request package, so this is just an emergency fix. This also lowers the default Protocol.SESSION_POOLSIZE to 1 because no internal code is running multi-threaded anymore.
• All-day calendar items (created as CalendarItem(is_all_day=True, ...)) now accept EWSDate instances for the start and end values. Similarly, all-day calendar items fetched from the server now return start and end values as EWSDate instances. In this case, start and end values are inclusive; a one-day event starts and ends on the same EWSDate value.
• Add support for RecurringMasterItemId and OccurrenceItemId elements that allow to request the master recurrence from a CalendarItem occurrence, and to request a specific occurrence from a CalendarItem master recurrence. CalendarItem.master_recurrence() and CalendarItem.occurrence(some_occurrence_index) methods were added to aid this traversal. some_occurrence_index in the last method specifies which item in the list of occurrences to target; CalendarItem.occurrence(3) gets the third occurrence in the recurrence.
• Change Contact.birthday and Contact.wedding_anniversary from EWSDateTime to EWSDate fields. EWS still expects and sends datetime values but has started to reset the time part to 11:59. Dates are a better match for these two fields anyway.
• Remove support for len(some_queryset). It had the nasty side-effect of forcing list(some_queryset) to run the query twice, once for pre-allocating the list via the result of len(some_queryset), and then once more to fetch the results. All occurrences of len(some_queryset) can be replaced with some_queryset.count(). Unfortunately, there is no way to keep backwards-compatibility for this feature.
• Added Account.identity, an attribute to contain extra information for impersonation. Setting Account.identity.upn or Account.identity.sid removes the need for an AD lookup on every request. upn will often be the same as primary_smtp_address, but it is not guaranteed. If you have access to your organization's AD servers, you can look up these values once and add them to your Account object to improve performance of the following requests.
• Added support for CBA authentication

• The max_wait argument to FaultTolerance changed semantics. Previously, it triggered when the delay until the next attempt would exceed this value. It now triggers after the given timespan since the first request attempt.
• Fixed a bug when pagination is combined with max_items (#710)
• Other minor bug fixes

• Removed the legacy autodiscover implementation.
• Added QuerySet.depth() to configure item traversal of querysets. Default is Shallow except for the CommonViews folder where default is Associated.
• Updating credentials on Account.protocol after getting an UnauthorizedError now works.

• The new Autodiscover implementation added in 2.2.0 is now default. To switch back to the old implementation, set the environment variable EXCHANGELIB_AUTODISCOVER_VERSION=legacy.
• Removed support for Python 2

• Added support for specifying a separate retry policy for the autodiscover service endpoint selection. Set via the exchangelib.autodiscover.legacy.INITIAL_RETRY_POLICY module variable for the the old autodiscover implementation, and via the exchangelib.autodiscover.Autodiscovery.INITIAL_RETRY_POLICY class variable for the new one.
• Support the authorization code OAuth 2.0 grant type (see issue #698)
• Removed the RootOfHierarchy.permission_set field. It was causing too many failures in the wild.
• The full autodiscover response containing all contents of the reponse is now available as Account.ad_response.
• Added a new Autodiscover implementation that is closer to the specification and easier to debug. To switch to the new implementation, set the environment variable EXCHANGELIB_AUTODISCOVER_VERSION=new. The old one is still the default if the variable is not set, or set to EXCHANGELIB_AUTODISCOVER_VERSION=legacy.
• The Item.mime_content field was switched back from a string type to a bytes type. It turns out trying to decode the data was an error (see issue #709).

• Bugfix release.

• Added support for OAuth 2.0 authentication
• Fixed a bug in RelativeMonthlyPattern and RelativeYearlyPattern where the weekdays field was thought to be a list, but is in fact a single value. Renamed the field to weekday to reflect the change.
• Added support for archiving items to the archive mailbox, if the account has one.
• Added support for getting delegate information on an Account, as Account.delegates.
• Added support for the ConvertId service. Available as Protocol.convert_ids().

• Fixed a bug where version 2.x could not open autodiscover cache files generated by version 1.x packages.

• Item.mime_content is now a text field instead of a binary field. Encoding and decoding is done automatically.

• The Item.item_id, Folder.folder_id and Occurrence.item_id fields that were renamed to just id in 1.12.0, have now been removed.

• The Persona.persona_id field was replaced with Persona.id and Persona.changekey, to align with the Item and Folder classes.

• In addition to bulk deleting via a QuerySet (qs.delete()), it is now possible to also bulk send, move and copy items in a QuerySet (via qs.send(), qs.move() and qs.copy(), respectively).

• SSPI support was added but dependencies are not installed by default since it only works in Win32 environments. Install as pip install exchangelib[sspi] to get SSPI support. Install with pip install exchangelib[complete] to get both Kerberos and SSPI auth.

• The custom extern_id field is no longer registered by default. If you require this field, register it manually as part of your setup code on the item types you need:

  from exchangelib import CalendarItem, Message, Contact, Task
  from exchangelib.extended_properties import ExternId
  
  CalendarItem.register("extern_id", ExternId)
  Message.register("extern_id", ExternId)
  Contact.register("extern_id", ExternId)
  Task.register("extern_id", ExternId)
  

• The ServiceAccount class has been removed. If you want fault tolerance, set it in a Configuration object:

  from exchangelib import Configuration, Credentials, FaultTolerance
  
  c = Credentials("foo", "bar")
  config = Configuration(credentials=c, retry_policy=FaultTolerance())
  

• It is now possible to use Kerberos and SSPI auth without providing a dummy Credentials('', '') object.

• The has_ssl argument of Configuration was removed. If you want to connect to a plain HTTP endpoint, pass the full URL in the service_endpoint argument.

• We no longer look in types.xsd for a hint of which API version the server is running. Instead, we query the service directly, starting with the latest version first.

• Bugfix release.

• Fix bug that left out parts of the folder hierarchy when traversing account.root.
• Fix bug that did not properly find all attachments if an item has a mix of item and file attachments.

• Add support for reading and writing PermissionSet field on folders.
• Add support for Exchange 2019 build IDs.

• Add Protocol.expand_dl() to get members of a distribution list.

• Lower the session pool size automatically in response to ErrorServerBusy and ErrorTooManyObjectsOpened errors from the server.
• Unusual slicing and indexing (e.g. inbox.all()[9000] and inbox.all()[9000:9001]) is now efficient.
• Downloading large attachments is now more memory-efficient. We can now stream the file content without ever storing the full file content in memory, using the new Attachment.fp context manager.

• Add a MAINFEST.in to ensure the LICENSE file gets included + CHANGELOG.md and README.md to sdist tarball
• Renamed Item.item_id, Folder.folder_id and Occurrence.item_id to just Item.id, Folder.id and Occurrence.id, respectively. This removes redundancy in the naming and provides consistency. For all classes that have an ID, the ID can now be accessed using the id attribute. Backwards compatibility and deprecation warnings were added.
• Support folder traversal without creating a full cache of the folder hierarchy first, using the some_folder // 'sub_folder' // 'leaf' (double-slash) syntax.
• Fix a bug in traversal of public and archive folders. These folder hierarchies are now fully supported.
• Fix a bug where the timezone of a calendar item changed when the item was fetched and then saved.
• Kerberos support is now optional and Kerberos dependencies are not installed by default. Install as pip install exchangelib[kerberos] to get Kerberos support.

• Improve back off handling when receiving ErrorServerBusy error messages from the server
• Fixed bug where Account.root and its children would point to the root folder of the connecting account instead of the target account when connecting to other accounts.

• Add experimental Kerberos support. This adds the pykerberos package, which needs the following system packages to be installed on Ubuntu/Debian systems: apt-get install build-essential libssl-dev libffi-dev python-dev libkrb5-dev.

• Bugfix release

• Bugfix release

• Added cancel to CalendarItem and CancelCalendarItem class to allow cancelling meetings that were set up

• Added accept, decline and tentatively_accept to CalendarItem as wrapper methods

• Added accept, decline and tentatively_accept to MeetingRequest to respond to incoming invitations

• Added BaseMeetingItem (inheriting from Item) being used as base for MeetingCancellation, MeetingMessage, MeetingRequest and MeetingResponse

• Added AssociatedCalendarItemId (property), AssociatedCalendarItemIdField and ReferenceItemIdField

• Added PostReplyItem

• Removed Folder.get_folder_by_name() which has been deprecated since version 1.10.2.

• Added Item.copy(to_folder=some_folder) method which copies an item to the given folder and returns the ID of the new item.

• We now respect the back off value of an ErrorServerBusy server error.

• Added support for fetching free/busy availability information ofr a list of accounts.

• Added Message.reply(), Message.reply_all(), and Message.forward() methods.

• The full search API now works on single folders and collections of folders, e.g. some_folder.glob('foo*').filter() , some_folder.children.filter() and some_folder.walk().filter().

• Deprecated EWSService.CHUNKSIZE in favor of a per-request chunk_size available on Account.bulk_foo() methods.

• Support searching the GAL and other contact folders using some_contact_folder.people().

• Deprecated the page_size argument for QuerySet.iterator() because it was inconsistent with other API methods. You can still set the page size of a queryset like this:

  qs = a.inbox.filter(...).iterator()
  qs.page_size = 123
  for item in items:
      print(item)
  

• Added support for registering extended properties on folders.
• Added support for creating, updating, deleting and emptying folders.