extract-msg

v0.55.0

• [TeamMsgExtractor #465] Added missing msg.close() to openMsg(). If the MSG file was actually just a plain OLE file, it would be left open.
• Adjusted the default value of maxNameLength for MessageBase.save() to 40 instead of 256.
• Adjusted exception handling for MessageBase.save() to properly report the reason a folder fails to be created.
• Simplified some of the code for MessageBase.save().
• Fixed some typing information.

v0.54.1

• [TeamMsgExtractor #462] Fix potential issue where child MSG might have incompatible encoding to parent MSG when trying to grab a stream from the parent.
• Added code to attempt to significantly improve RTF deencapsulation times. This tries to strip away unneeded data before passing it to RTFDE. This shows improvements on all files that take more than one second. Currently, this actually fixes some files previously outputting wrong from RTFDE when deencapsulating the HTML body, specifically around non breaking spaces sometimes not transferring over.

v0.54.0

• [TeamMsgExtractor #456] Changed the prepared html output to use plainly encoded HTML instead of prettified, since current prettification options used mangles the output and causes the output to sometimes be very large.

v0.53.2

• [TeamMsgExtractor #452] Adjusted code to allow html encoding to be cached to try to speed up bs4 operations.
• [TeamMsgExtractor #453] Fixed handler for too large filetimes so that some filetimes being too large doesn't break the handler.
• Fixed a bug that would cause an error in task objects due to a lack of enumerate.
• Fix TOCEntry not initializing DVTargetDevice correctly.
• Add temporary properties for ContentID to SignedAttachment. AFAIK these can't ever be set, but this prevents errors in some places.

v0.53.1

• Expanded allowable range for red-black-tree-mod.
• Fix issue with MessageBase.asEmailMessage() that prevented embedded MSG files from being attached.
• Expand allowable versions of BeautifulSoup4.

v0.52.0

• [TeamMsgExtractor #444] Fix typo in string that prevented HTML body from generating from the plain text body properly.
• Adjusted the behavior of MSGFile.areStringsUnicode to prioritize the property specified by the parent MSG files for MSG files that are embedded. Additionally, added a fallback to rely on whether or not there is a stream using the 001F type to determine the property value if it is entirely missing.
• Adjusted OleWriter.fromMsg() and MSGFile.export() to add the argument allowBadEmbed which helps to correct a few different issues that may appear in embedded MSG files. These corrections allow the embedded file to still be extracted and to open properly in Outlook.
• In addition to the above, the errors that some of those corrections will suppress are now significantly more informative about what went wrong.

v0.51.1

• Add class type added in last version to known class types.

v0.51.0

• [TeamMsgExtractor #401] Add basic support for MSG class type IPM.SkypeTeams.Message.

v0.50.1

• [TeamMsgExtractor #434] Fix bug introduced in previous version.

v0.50.0

• [TeamMsgExtractor #432] Adjust html header code to replace non-ascii characters with escaped versions. Also adujusted plain text to html conversion to ensure non-ascii character from the body are encoded to escpaed values to be safe.
• Made some corrections to NullDate.

v0.49.0

• [TeamMsgExtractor #427] Adjusted code for converting time stamps to create null dates for any time stamp beyond a certain point. The point was determined to be close to the existing null dates.
• [TeamMsgExtractor #425] Added basic support for custom attachments that are Windows Metafiles.
• Changed tolerance of bitmap custom attachment handler to allow for attachments with only a CONTENT stream. This change was made after seeing an example of a file that only had a CONTENT stream and no other streams for the custom data. The code now also tries to create default values for things previously determined from those other streams.
• Fixed an issue in tryGetMimetype were the code didn't properly check if the data type was bytes (it only checked if it had a type).
• Corrected some exports.
• Added new ErrorBehavior value CUSTOM_ATTACH_TOLERANT to allow skipping checks for unused data that is normally validated.

v0.48.7

• [TeamMsgExtractor #420] Fixed typo introduced in last version.

v0.48.6

• [TeamMsgExtractor #417] Fixed issues with openMsg where some corrupted MSG files could end up throwing an uncaught exception and leaving the file handle open.

v0.48.5

• [TeamMsgExtractor #414] Fixed typo in message_signed_base.py.

v0.48.4

• [TeamMsgExtractor #411] Fix console script throwing error due to changed console args not defaulting.

v0.48.3

• [TeamMsgExtractor #409] Added missing private method to SignedAttachment.
• Fixed some missing typing information.

v0.48.2

• Fixed bugs with MessageBase.asEmailMessage(). Numerous improvements to how it handles the data.

v0.48.1

• Added an option (-s, --stdin) to the command line to take an MSG file from stdin. This allows the user to pipe the MSG data from another program directly instead of having to write a middleman that uses the extract-msg library directly or having to write the file to the disk first.
• Changed main function to allow for manual argument list to be passed to it.
• Added attributes to AttachmentBase for creation and modification time. These can be accessed through createdAt or creationTime and lastModificationTime or modifiedAt.
• Changed OleWriter tests to output the name of the test file being done if an error occurs.
• Added tests for some command line stuff.

v0.48.0

• Adjusted error handling for named properties to handle critical streams being missing and to allow suppression of those errors.
• Adjusted error handling for named properties to allow silencing of errors caused by invalid references to the name stream. If ErrorBehavior.NAMED_NAME_STREAM is provided to the MSGFile instance, a warning will be logged and that entry will simply be dropped.
• Adjusted error handling for signed messages to better check for issues with the signed attachment. This should make errors from violating the standard much easier to understand. These errors can be ignored, but the attachment will not be parsed as a signed attachment.
• Minor docstring updates.
• Minor adjustments to OleWriter to prepare the code for being able to write version 4 files. Version 3 files are currently the only one's supported, but much of the code had hard-coded values that could be replaced with variables and small conditionals. This will have very little performance impact, and should not be noticeable.
• Improved comments on OleWriter to make private sections more understandable.
• Changed MessageSignedBase._rawAttachments to MessageSignedBase.rawAttachments to provide non-private access in a reliable way.

v0.47.0

• Changed the public API for PropertiesStore to improve the quality of its code. The properties type is now mandatory, and the intelligence field (and the related enum) has been removed.
  • Additionally, the toBytes and __bytes__ methods will both generate based on the contents of this class, allowing for new properties to be created and for existing properties to be modified or removed if the class is set to writable on creation.
• Added new method Named.getPropNameByStreamID. This method takes the ID of a stream (or property stream entry) and returns the property name (as a tuple of the property name/ID and the property set) that is stored there. Returns None if the stream is not used to store a named property. This name can be directly used (if it is not None) to get the NamedPropertyBase instance associated. This method is most useful for people looking at the raw data of a stream and trying to figure out what named property it refers to.
• Fixed mistake in struct definitions that caused a float to require 8 bytes to unpack.
• Added tests for extract_msg.properties.props.
• Added basic tests for extract_msg.attachments.
• Added validation tests for the enums and constants submodule.
• Removed unneeded structs.
• Fixed an issue where PtypGuid was being parsed by the wrong property type. Despite having a fixed size, it is still a variable length property.
• Fixed all of the setters not working. I didn't know they needed to use the same name as the getter, and I swear they were working at some point with the current setup. Some sources online suggested the original form should work, which is even stranger.
• Unified all line endings to LF instead of a mix of CRLF and LF.
• Changed enums BCTextFormat and BCLabelFormat to IntFlag enums. The values that exist are for the individual flags and not the groups of flags.
• Made FieldInfo writable, however it can no longer be directly converted to bytes since it requires additional information outside of itself to convert to bytes. It still retains a toBytes method, however it requires an argument for the additional data.
• Fixed UnsupportedAttachment inverting the skipNotImplemented keyword argument.
• Fixed DMPaperSize not being a subclass of Enum.
• Extended values for DMPaperSize.
• Removed unneeded structs.
• Fixed exports for extract_msg.constants.st.
• Updated various parts of the documentation to improve it and make it more consistent.
• Fixed Recipient.existsTypedProperty and AttachmentBase.existsTypedProperty having the wrong return type.
• Removed "TODO" markers on OleStreamStruct and finalized it to only handle the OLEStream for embedded objects.
• Fixed type annotations for extract_msg.utils.fromTimeStamp.
• Added new function extract_msg.properties.prop.createNewProp. This function allows a new property to be created with default data based on the name. The name MUST be an 8 character hex string, the first 4 characters being the value for the property ID and the second 4 being the value for the type.
• Fixed VariableLengthProp.reservedFlags returning the wrong type.
• Adjusted many of the property structs to make it easier to use them for packing.
• Renamed some of the property structs to be more informative.
• Fixed ServerID using the wrong struct (would have thrown an exception for not having enough data).
• Unified signing for integer properties. All integer properties are considered unsigned by default when read directly from the MSG file, however specific property accessors may convert to signed if the property is specifically intended to be so. This does not necessarily include properties that are stored as integers but have a value that is not an integer, like PtypCurrency.
• Changed extract_msg.constants.NULL_DATE to be a subclass of datetime.datetime instead of just a fixed value. Functions that return null dates may end up returning distinct versions of NullDate (the newly created subclass), however all instances of NullDate will register as being equal to each other. Existing equality comparisons to the NULL_DATE constant will all function as intended, however is checks may fail for some null dates.
• Changed currency type to return a Decimal instance for higher precision.
• Made FixedLengthProp and VariableLengthProp writable. Only the property flags from PropBase is writable. This also includes the ability to convert them to bytes based on their value.
• Fixed many issues in VariableLengthProp regarding it's calculation of sizes.
• Removed hasLen function.
• Changed style to remove space between variable name and colon that separates the type.
• Corrected InvaildPropertyIdError to InvalidPropertyIdError.
• Updated to olefile version 0.47.
• Updated RTFDE minimum version to 0.1.1.
• Changed dependency named for compressed-rtf to remove minor typo (it should forward to the correct place regardless, but just to be safe).
• Fixed issues with implementation of OleWriter when it comes to large sets of data. Most issues would fail silently and only be noticeable when trying to open the file. If your file is less than 2 GB, you would likely not have notices and issues at all. This includes adding a new exception that is throw from the write method, TooManySectorsError.
• Fixed some issues with OleWriter that could cause entries to end up partially added if the data was an issue.

v0.46.2

• Adjusted typing information on regular expressions. They were using a subscript that was added in Python 3.9 (apparently that is something the type checker doesn't check for), which made the module incompatible with Python 3.8. If you are using Python 3.9 or higher a version check will switch to the more specific typing.

v0.46.1

• [TeamMsgExtractor #394] Fix typo in props that caused the wrong number of bytes to be given to a struct.

v0.46.0

• [TeamMsgExtractor #95] Adjusted the overrideEncoding property of MSGFile to allow automatic encoding detection. Simply set the property to the string "chardet" and, assuming the chardet module is installed, it will analyze a number of the strings to try and form a consensus about the encoding. This will ignore the specified encoding only if if successfully detects. Otherwise it will log a warning and fall back to the default behavior.
• [TeamMsgExtractor #387] Changed extract_msg.utils.decodeRfc2047 to not throw decoding errors if the content given is not ASCII.
• [TeamMsgExtractor #387] Changed header parsing policy to email.policy.compat32 to prevent partial parsing of quoted header fields.
• [TeamMsgExtractor #388] Updated documentation of MSGFile.export to specify that updated fields on an MSGFile instance (and it's subclasses) will not be reflected in the result of the function. Many of the functions do use the newest version of a cached_property, but this is not one of them.
• Removed methods deprecated in v0.45.0.
• Changed the base class of EntryID from no base class to abc.ABC.
• Added position property to EntryID to tell how many bytes were used to create the EntryID.
• Added a number of properties to MSGFile from [MS-OXCMSG].
• Moved some properties down to MessageBase from it's subclasses.
• Added support for Journal objects.
• Changed internal code of PermanentEntryID to correctly parse the data. Previously the distinguished name did not actually end at the null character, instead ending at the end of the bytes provided. If there was trailing data, it would be captured inadvertently.
• Finished definition for StoreObjectEntryID.
• Added new keyword arguments for MSG files: dateFormat and datetimeFormat. These allow the user to easily override the strings being used for format dates and dates that include a time component, respectively.
  • In unifying all the formats into 2 options, you may notice that some will look a bit different starting from this version, as there was an unfortunately large amount of variation.
• Fixed code for MessageBase.parsedDate which could have incorrect values.
• Fixed issues with MessageBase.date and related things either being incorrectly documented or doing things that are not specified by the documentation. It was supposed to have been changed to use datetime objects, but it was still using strings.
• Removed unused function extract_msg.utils.isEmptyString.
• Removed unused function extract_msg.utils.properHex.
• Added a getStream helper function to CustomAttachmentHandler.
• Added a getStreamAs helper function to CustomAttachmentHandler.
• Added new custom attachment handler for journal-associated attachments.
• Changed EntryID.autoCreate to return None if given None or empty bytes.
• Changed EntryID.autoCreate to raise a FeatureNotImplemented exception if no valid entry ID class is found.
• Fix typing annotations for CustomAttachmentHandler.
• Removed unneeded imapclient dependency.
• Changed getJson to have values be null if they aren't found rather than an empty string.
• Implemented the getJson method correctly for a number of classes.
• Changed Task.percentComplete to always return a float.
• Changed the NotImplementedError for custom attachment handler not being found to FeatureNotImplemented. Additionally, changed the error message to specify the CLSID found on the attachment to better enable people to report issues.
• Changed code for Recipient and MessageBase that makes it rely on MessageBase.recipientTypeClass to determine the class to use for the recipientType property. Adjusted the typing of Recipient to have it reflect the type that will be used.
• Correctly changed the returned value for ResponseStatus.fromIter to actually return a List instead of a set.
• Filled out typing information for a significant portion of the module where variables or functions were missing it. This includes the entirety of the constants submodule.
• Corrected a number of minor issues.
• Extended values for DVAspect enum.
• Added new enums to go with parsing for OLEPresentationStream.
• Changed NNTPNewsgroupFolderEntryID.newsgroupName to bytes instead of string since it is ANSI.
• Fixed an issue that would cause headers to fail to parse properly if the header text starts with "Microsoft Mail Internet Headers Version 2.0" which is common on some MSG files. This is fixed by stripping that from the beginning before actually parsing the text. This is to circumvent CPython issue #85329, confirmed to still exist in at least some of the supported Python versions.
• Added listDir and slistDir as methods to AttachmentBase, Recipient, and Named. These always exclude the prefix, returning as if their directory is the root of the object. This allows the named to be directly used for accessing those files.
• Numerous spelling fixes in docstrings, comments, and exceptions.
• Reduced the amount of initialization performed by MessageBase. Much of this initialization was there from before a lot of stuff changed to cached_property and a number of internal variables were being used. Now all of the relevant variables will be initialized by the way they are accessed.
• Added new exception DependencyError.
• Changed the errors for missing optional dependencies from ImportError to DependencyError.
• Removed all instances of the rawData property in favor of the toBytes method. For now, many of these will simply return the raw data used, specifically those that are still unmodifiable. Any whose properties have the ability to be modified will have properly implemented versions. These classes also allow None to be passed as the value for their data, which will be the default if no arguments have been passed to the constructor. If no arguments or None is given as the data, it will create a new instance with default values. This is all in an effort to move towards the ability to create new MSG files and the MSGWriter class. All toBytes methods will either exclusively return bytes or will return None to specify that the structure isn't valid to convert to bytes. Structures that may be invalid will be annotated as Optional[bytes] for the return type.
  • Additionally, these objects will also support the __bytes__ method. If the object returned is not bytes, the method will throw a TypeError.
• Removed the individual PropBase flag properties and changed the main flags property to return an enum containing the flags.
• Changed various data structs to allow modification and creation of new instances for writing to an MSG file.
• Changed TZRule to use unsigned values where applicable.
• Changed TZRule to require the 14 null bytes (I commented it out completely on accident instead of swapping it to a plain read). It now logs a warning about the bytes not being null.
• Removed unneeded function windowsUnicode.
• Moved FixedLengthProperty.parseType to the private API. This was not intended for external use anyways, so leaving it as public API didn't make sense.
• Fixed check for type in ContactAddressEntryID being the wrong value.
• Modified inputToBytes to support objects with the __bytes__ method. If the method exists and works then it will be used as a last resort.
• Modified OleWriter to accept objects with a __bytes__ method for the data to use for an entry.
• Added __bytes__ method to MSGFile. This is equivalent to calling MSGFile.exportBytes.

v0.45.0

• BREAKING: Changed parsing of string multiple properties to remove the trailing null byte. This will cause the output of parsing them to differ.
• Updated typing information for some functions and classes.
• Fixed a bug with MessageSignedBase.attachments that would cause it to return None instead of an empty list if the number of normal attachments was 0 was the error behavior was set to ignore violations of the standard.
• Updated MessageSignedBase.attachments to use functools.cached_property instead of property.
• Fixed spelling errors in some exception strings.
• Made NamedPropertyBase a subclass of abc.ABC.
• Cleaned up some of the code for named properties to remove unused variables and remove inefficient code.
• Changed PropBase to be a subclass of abc.ABC.
• Added detailed versioning info to the README.
• Deprecated many private functions, including methods on many of the classes. Of primary note are _getStream and _getStringStream, which have been moved to the public API as getStream and getStringStream. Any deprecated functions still exist and will forward to a public API function if they are not being removed. Additionally, all internal usage of them has been removed. This change is one of the big preparations that is needed for the 1.0.0 release.
  • As mentioned, a number of these deprecated functions have been moved to the public API. It is recommended that you run tests with your code after enabling deprecation warnings to see what should be changed.
• Removed items deprecated in or before 0.42.0.
• Changed the API for the private method _genRecipient. This is not intended for use outside of the module except for subclasses. The change removed the allowance of ints for the second argument, requiring that it be a valid enum type.
• Convert many enum types to IntEnum.
• Extended functionality of PropertiesStore to allow for integer property names and getting a property based on just the ID. You can also get a list of all properties that use a given ID.
• Added new function PropertiesStore.getProperties which gets a list of all properties matching the property ID. Return type is a list of PropBase instances.
• Added new function PropertiesStore.getValue which looks for the first matching FixedLengthProp and returns the value from it.
• Improved internal code related to getting a property with a potentially unknown type.
• Added a number of entirely new functions to the public API on MSGFile, AttachmentBase, PropertiesStore, and Recipient objects:
  • getMultipleBinary: Gets a multiple binary property as a list of bytes objects.
  • getSingleOrMultipleBinary: A combination of getStream and getMultipleBinary which prefers a single binary stream. Returns a single bytes object or a list of bytes objects.
  • getMultipleString: Gets a multiple string property as a list of str objects.
  • getSingleOrMultipleString: A combination of getStringStream and getMultipleString which prefers a single string stream. Returns a single bytes objecct or a list of bytes objects.
  • getPropertyVal: Shortcut for instance.props.getValue that allows new behavior to be added by overriding it.
  • getNamedProp: Shortcut for instance.namedProperties.get((propertyName, guid), default) that allows new behavior to be added by overriding it.
• Removed Named._getStringStream and Named.sExists. The named properties storage will always use regular streams and not string streams.
• Changed all Named methods to no longer have a prefix argument. The prefix should always be false sense the named property mapping will only exist in the top level directory.
• Adjusted tryGetMimeType to allows any attachments whose data property would return a bytes instance.
• Changed internal code to use public API functions wherever possible. This includes making many private API functions use calls to the public API for getting bits of data.
• Fixed potential issue with AttachmentBase.clsid which had the potential to cause some attachments to fail to generate a CLSID.
• Outright removed or changed a significant portion of the private API. I have rarely, if ever, seen references to these parts, so this should cause you no issues. Some of these have also been moved to the public API, either identically or with changes, and the mapping is as such:
  • _getNamedAs -> getNamedAs: Changed to always require a conversion argument. If you were previously using it to plainly get a named property or to handle the properly being None or a real value, you should use the return value of getNamedProp instead.
  • _getPropertyAs -> getPropertyAs: Same as above, use getPropertyVal instead for None or plain access.
  • _getStreamAs -> getStreamAs, getStringStreamAs: Once again, see above. Use getStream and getStringStream, respectively.

v0.44.0

• Fixed a bug that caused MessageBase.headerInit to always return False after the 0.42.0 update.
• Changed MessageBase.headerInit to a property.
• Fixed extract_msg.utils.__all__.
• Minor regoanization within extract_msg/utils.py.
• Minor changes to docstrings.
• Minor README updates.
• Fix issue with folded header fields decoding incorrectly when given to extract_msg.utils.decodeRfc2047.

v0.43.0

• [TeamMsgExtractor #56] [TeamMsgExtractor #248] Added new function MessageBase.asEmailMessage which will convert the MessageBase instance, if possible, to an email.message.EmailMessage object. If an embedded MSG file on a MessageBase object is of a class that does not have this function, it will simply be attached to the instance as bytes.
• Changed imports in message_base.py to help with type checkers.
• Changed from using email.parser.EmailParser to email.parser.HeaderParser in MessageBase.header.
• Changed some of the internal code for MessageBase.header. This should improve usage of it, and should not have any noticeable negative changes. You man notice some of the values parse slightly differently, but this effect should be mostly suppressed.

v0.42.2

• Fix bug in AttachmentBase.mimetype that would cause it to throw an error when accessed. This bug was introduced in v0.42.0.

v0.42.1

• Fixed some constants being accessed with the wrong name (names were changed in reorganization).
• Removed unused regular expression.

v0.42.0

• [TeamMsgExtractor #372] Changed the way that the save functions return a value. This makes the return value from all save functions much more informative, allowing a user to separate if a file or folder (or if more than one) was saved from the function. It also guarantees that all classes from this module will return the relevant path(s) if data is actually saved.
• [TeamMsgExtractor #288] Added feature to allow attachment save functions to simply overwrite existing files of the same name. This can be done with the overwriteExisting keyword argument from code or the --overwrite-existing option from the command line.
• [TeamMsgExtractor #40] Added new submodule custom_attachments. This submodule provides an extendable way to handle custom attachment types, attachment types whose structure and formatting are not defined in the Microsoft documentation for MSG files. This includes a handler to at least partially cover support for Outlook images.
• [TeamMsgExtractor #373] Added the encoding submodule for encoding tasks, including proper support for Microsoft's implementation of CP950. This gets added to the codecs list as "windows-950".
  • Added infrastructure to make it easy to add variable-byte (up to two bytes) encodings and single-byte encodings.
  • Added the following encodings:
    • windows-874
    • x-mac-ce
    • x-mac-cyrillic
    • x-mac-greek
    • x-mac-icelandic
    • x-mac-turkish
• Fixed an issue in the save functions that left the possibility for the zip files to not end up closing if the save function created it and then had an exception.
• Added new property AttachmentBase.clsid which returns the listed CLSID value of the data stream/storage of the attachment.
• Changed internal behavior of MSGFile.attachments. This should not cause any noticeable changes to the output.
• Refactored code significantly to make it more organized.
• Changed the exports from the main module to only include an important subset of the module. For other items, you'll have to import the submodule that it falls under to access it. Submodules export all important pieces, so it will be easier to find.
  • This includes having many modules be under entirely new paths. Some of these changes have been done with no deprecation, something I generally try to avoid. This is happening at the same time as the public api is significantly changing, which makes it more acceptable.
• Fixed __main__ using the wrong enum for error behavior.
• Fixed Named.get being severely out of date (it's not used anywhere by the module which is why it wasn't noticed).
• Fixed Named.__getitem__ being entirely case-sensitive.
• Switched much of the internal code (and the treePath property of all classes that have it) to using weakref.ReferenceType to avoid hard cyclic references.
• Fixed Recipient._getTypedStream never returning a value.
• Added additional type hints in various places.
• Modified tests.py to only run if it is run as a file instead of imported.
• Changed knownMsgClass to a private function since it is explicitly not being exported by any part of the module.
• Removed unused function getFullClassName.
• Fixes to the HTML body when saving as HTML will no longer require the preparedHtml/--prepared-html option.
• Removed unused exceptions.
• Entirely reoganized the way attachments are initialized, including the class that will be used in various circumstances. Embedded MSG files, custom attachments, and web attachments will all use dedicated classes that are subclasses of AttachmentBase.
  • With this change, the way to specify a new Attachment class is to override the function used when creating attachments. This can be done by passing attachmentInit = myFunction as an option to openMsg. This function MUST return an instance of AttachmentBase.
• Added first implementation of web attachments. Saving is not currently possible, but basic relevant property access is now possible. Saving will not be stopped by this attachment if skipNotImplemented = True is passed to the save function.
• Changed the option to suppress RTFDE errors to fall under the ErrorBehavior enum. Usage of the original option will be allowable, but is being marked as deprecated. However, it is still a dedicated option from the command line.
  • Also fixed the option not properly ignoring some RTFDE errors, specifically the ones that it is normal for the module to throw.
• Removed some constants that are not used by the module.
• Updated to support RTFDE version 0.1.0. Users encountering random errors from that module should find that those errors have disappeared. If you get errors from it still, bring up the issue on their GitHub.
• Fixed bug that would cause weird behavior if you gave an empty string as the path for an MSG file.
• Added support for IPM.StickyNote.
• Fixed an issue that would cause MSG file to never close if an error happened during any of the __init__ functions for MSG classes.
• Removed unneeded chardet dependency.
• Removed Contact.__init__ as it didn't provide any unique behavior.
• Changed the documentation of openMsg to specify that it accepts all options recognized by MSGFile subclasses, allowing the doc string to not be modified every time one of them is changed.
  • Changed the documentation of various __init__ methods to do the same thing.
• Added dataType property to AttachmentBase and SignedAttachment for checking the class that the data will be, if accessible. Returns None if the data is inaccessible, including because accessing it would throw an exception.
• Added new enum InsecureFeatures and option insecureFeatures. This option will allow certain features with security implications to be used for files that you trust. Currently the only feature it supports is the usage of PIL/Pillow to open and modify images. All features like this will be opt-in to reduce possible vulnerabilities.
• Modified all custom exceptions the module uses to derive from a single base class for better organization.
  • Added new exceptions to handle some of the situations previously handled by base Python exceptions.
• Changed internal handling of the prefix option for MSGFile.__init__ (and therefore openMsg). If you are not setting this manually, you should notice little difference.
• Made enums less strict and converted all using fromBits to be IntFlag enums.
• Fixed CalendarBase.keywords being blatantly incorrect (it was so bad I don't know how it slipped through).
• Fixed Contact.gender being blatantly incorrect.
• Fixed sender not being properly decoded in some circumstances.
• Changed behavior of MSGFile to have olefile raise defects of type DEFECT_INCORRECT and above instead of just DEFECT_FATAL. Uncaught issues of DEFECT_INCORRECT can often cause the module to have parsing issues that may be misleading, this just ensures the issue is clarified. This behavior can be reverted back to the previous with ErrorBehavior.OLE_DEFECT_INCORRECT.
• Fixed potential issues that may have made is possible for certain attachments to ignore filename conflict resolution code.

v0.41.5

• Fixed an issue from version 0.41.3 where the header being present but missing the From field would cause an exception.

v0.41.4

• Fixed an issue in the last version that would break the decoding function if the contents were not encoded.
• Updated tzlocal and allow future updates for compressed_rtf and ebcdic.

v0.41.3

• [TeamMsgExtractor #365] Fixed an issue that would cause certain values retrieved from the header to not be decoded properly. It does this when retrieving the values, so nothing about the header has been changed.
• Added new property MessageBase.headerText which is the text content of the header stream. Adjusted other things to use this instead of trying to retrieve the stream directly in multiple places.
• Added typing to MessageBase.header.

v0.41.2

• Updated annotations on MessageBase.save.
• Added new enum BodyTypes.
• Added property MessageBase.detectedBodies for detecting what bodies have been stored (not generated by the module) in the .msg file.

v0.41.1

• [TeamMsgExtractor #362] Fixed an issue with the removal of the --dev option missing one of the checks (I swear I actually tested it).

v0.41.0

• [TeamMsgExtractor #357] Fixed an issue where the properties stream being absent would raise an error message that was not clear.
• [TeamMsgExtractor #357] Added a way to suppress StandardViolationError for poorly created files. This may cause issues you might not expect since this exception is meant to stop the processing for a reason.
• [TeamMsgExtractor #223] Finally got around to dealing with signed attachments that are embedded MSG files. SignedAttachment.data now returns either bytes or MSGFile. SignedAttachment also now has a asBytes property that will return the bytes that created the signed attachment, regardless of if it is an MSG or not, making it unnecessary to call MSGFile.exportBytes to get the bytes of the embedded MSG file, which can add a small delay to your code. As Attachment is a much more complex class, it does not (at least yet) have this property. Also unlike Attachment, SignedAttachment will not throw an exception if the data is an MSG file but is not supported. Instead, it will simply be logged as a exception, but the code will continue. If the data is successfully read as an embedded MSG file, the AttachmentType will be AttachmentType.SIGNED_EMBEDDED.
• Deprecated AttachErrorBehavior in favor of the new ErrorBehavior enum which controls the error behavior for the various parts of the MSG file. Uses of the former will work until the next major version.
• Deprecated the attachmentErrorBehavior parameter of MSGFile in favor of errorBehavior. Uses of the previous will work until the next major version.
• Added treePath property to SignedAttachment to bring it more in line with AttachmentBase.
• Fixed bug in rare logging message caused by incorrect type name.
• Removed the dev and dev_classes submodules, as most of their features are possible using the base code. Additionally, the classes involved because significantly outdated over time.
• Removed the --dev argument from the command line.
• Changed the message for the standards violation error when an attachment has no specified attachment type and it could not be determined automatically. The error now specifies the path of the attachment that had an issue for easier debugging. Additionally, the log message for it simply not being present has also been changed to show this information.
• Added a dev level log that will output the entire properties mapping if the attachment type is not set. Dev level is 5.
• Fixed a critical error in Properties that caused the __contains__ method to always be False. This occurred because it was missing a return statement. Fortunately, it looks like only one part of the module was affected due to other parts using a properly written function.
• Removed some debug prints that slipped through.
• Changed some parts to use in for checking that a property exists as opposed to has_key. The function was there to act more like Python 2.
• Deprecated Properties.has_key.
• Removed the validation submodule and all related references. It was pretty outdated and has minimal usage at this point in time. It may come back at some later point.
• Changed behavior of MessageBase.save so it doesn't save raw when an exception occurs. This behavior may have ended up creating unexpected output which is why it was removed. It was mainly there for debugging in the first place, but is no longer necessary.
• Added __contains__ function to Named class.
• Fixed missing import in message_base.py that would only cause problems if something was wrong with the HTML.
• Fixed a bad error message appearing when trying to add or use an entry in OleWriter using an empty path.
• Changed the InvalidFileFormatError for a missing property stream to a StandardViolationError.
• Changed the exception messages for a few exceptions to fix typos and clarify.
• Fixed issues in _rtf.tokenize_rtf which would cause an exception to handle incorrectly and throw an unclear error.
• Fixed various small bugs caused by typos.
• Clean up unneeded imports.
• Improved existing __all__ entries and added some where they should be.
• Removed default export of the UnrecognizedMSGTypeError exception in favor of exporting the exceptions module.
• Removed default export of properHex.
• Improved type checking in many places.
• Fixed issues in RecurrencePattern.

v0.40.0

• [TeamMsgExtractor #338] Added new code to handle injection of text into the RTF body. For many cases, this will be much more effective as it relies on ensuring that it is in the main group and past the header before injection. It is not currently the first choice as it doesn't have proper respect for encapsulated HTML, however it will replace some of the old methods entirely. Solving this issue was done through the use of a few functions and the internal _rtf module. This module in it's entirety is considered to be implementation details, and I give no guarantee that it will remain in it's current state even across patch versions. As such, it is not recommended to use it outside of the module.
• Changed MessageBase.rtfEncapInjectableHeader and MessageBase.rtfPlainInjectableHeader from str to bytes. They always get encoded anyways, so I don't know why I had them returning as str.
• Updated minimum Python version to 3.8 as 3.6 has reached end of support and 3.7 will reach end of support within the year.
• Updated information in README.

v0.39.2

• Fixed issues with AttachmentBase.name that could cause it to generate wrong.
• Added convenience function MSGFile.exportBytes which returns the exported version from MSGFile.export as bytes instead of writing it to a file or file-like object.

v0.39.1

• [TeamMsgExtractor #333] Fixed typo in a warning.
• [TeamMsgExtractor #334] Removed __del__ method from MSGFile. It was there for cleanup, but wasn't planned well enough to stop it from causing issues. It may be reintroduced in the future if I can manage to remove the issues.
• [TeamMsgExtractor #335] Fixed some parts of extract_msg.utils.getCommandArgs having invalid logic after a previous (rather old) update that caused exceptions when using certain options.
• Added new property treePath to AttachmentBase and MSGFile (which adds it to nearly every class). This property is the path to the current instance, represented as a tuple of instances that would be used to get to the current instance.
• Added sphinx documentation.
• Fixed an issue in OleWriter that would produce corrupted OLE files if the DIFAT needed more than the header.

This release fixed several issues while also significantly increasing the functionality of the OleWriter class.

v0.39.0

• [TeamMsgExtractor #318] Added code to handle a standards violation (from what I can tell, anyways) caused by the attachment not having an AttachMethod property. The code will log a warning, attempt to detect the method, and throw a StandardViolationError if it fails.
• [TeamMsgExtractor #320] Changed the way string named properties are handled to allow for the string stream to have some errors and still be parsed. Warnings about these errors will be logged.
• [TeamMsgExtractor #324] Fixed an issues with contact saving when a list property returns None.
• [TeamMsgExtractor #326] Fixed a bug that could cause some files to error when exporting.
• Fixed an issue where creation and modification times were not being copied to the new OLE file created by OleWriter.
• Fixed up a few docstrings.
• Fixed a few issues in MSGFile regarding the filename keyword argument.
• Added new argument rootPath to OleWriter.fromOleFile for saving a specific directory from an OLE file instead of just copying the entire file. That directory will become the root of the new one.
• Adjusted code for OleWriter to generate certain values only at save time to make them more dynamic. This allows for existing streams to be properly edited (although has issues with allowing storages to be edited).
• Added new function OleWriter.deleteEntry to remove an entry that was already added. If the entry is a storage, all children will be removed too.
• Added new function OleWriter.editEntry to edit an entry that was already added.
• Added new function OleWriter.addEntry to add a new entry to the writer without an OleFileIO instance. Properties of the entry are instead set using the same keyword arguments as described in OleWriter.editEntry.
• Changed _DirectoryEntry to DirectoryEntry to make the more finalized version public. Access to the originals that the OleWriter class creates should never happen, instead copies should be returned to ensure the behavior is as expected.
• Added new function OleWriter.getEntry which returns a copy of the DirectoryEntry instance for that stream or storage in the writer. Use this function to see the current internal state of an entry.
• Added new function OleWriter.renameEntry which allows the user to rename a stream or storage (in place). This only changes it's direct name and not it's location in the new OLE file.
• Added new function OleWriter.walk which is similar to os.walk but for walking the structure of the new OLE file.
• Added new function OleWriter.listItems which is functionally equivalent to olefile.OleFileIO.listdir which returns a list of paths to every item. Optionally a user can get the paths just for streams, just for storages, or both. Requesting neither will simply return an empty list. Default is to just return streams.
• Added a small amount of path validation to inputToMsgPath which is used in a lot of places where user input for a path is accepted. It ensures illegal characters don't exist and that the path segments (each name for a storage or stream) are less than 32 characters. This will be most helpful for OleWriter.
• Added many internal helper functions to OleWriter to make extensions easier and consolidate common code. Many of these involve direct access to internal data which is why they are private.

v0.38.4

• Fix line in OleWriter that was causing exporting to fail.
• Fixed some issues with the README.

v0.38.3

• Fixed issues in HTML generation that caused line breaks to be omitted from large sections of the text.
• Fixed issues with requirements file.

v0.38.2

• Fixed new NameError accidentally introduced in the previous version.

v0.38.1

• Added a __del__ method to MSGFile to ensure a bit of proper cleanup should all references to an MSGFile instance be removed before the file is closed. OleFileIO doesn't appear to have one, so it's up to us to ensure it is properly closed. Note that the del keyword does not guarantee the immediate deletion of the object, and you should take care to close the file yourself. If this is not possible, importing the gc module and using it's collect method will free the files if your code has no references to them.
• Fixed an import issue with signed messages.

While the update doesn't add many features, it adds one of the features that has been requested since about the time I joined the module 4 years ago: The ability to extract embedded MSG files as .msg files. This is one of the more important milestones.

v0.38.0

• [TeamMsgExtractor #117] Added class OleWriter to allow the writing of OLE files, which allows for embedded MSG files to be extracted.
• Added function MSGFile.export which copies all streams and storages from an MSG file into a new file. This can "clone" an MSG file or be used for extracting an MSG file that is embedded inside of another.
• Added hidden function to MSGFile for getting the OleDirectoryEntry for a storage or stream. This is mainly for use by the OleWriter class.
• Added option extractEmbedded to Attachment.save (--extract-embedded on the command line) which causes embedded MSG files to be extracted instead of running their save methods.
• Fixed minor issues with utils.inputToMsgPath (renamed from utils.inputToMsgpath).
• Renamed utils.msgpathToString to utils.msgPathToString.
• Made some of the module requirements a little more strict to better version control. I'll be trying to make periodic checks for updates to the dependency packages and make sure that new versions are compatible before changing the allowed versions, while also trying to keep the requirements a bit flexible.

v0.37.1

• Added option to save function (including MSGFile.saveAttachments) to skip any attachments marked as hidden. This can also be done from the command line with --skip-hidden.
• Fixed an issue that could cause an MSGFile to be included in the attachment names instead of just strings.
• Fixed up several comments.
• Fixed bug that would cause spaces at the end of the subject or filename to break the module.

v0.37.0

• [TeamMsgExtractor #303] Renamed internal variable that wasn't changed when the other instances or it were renamed.
• [TeamMsgExtractor #302] Fixed properties missing (a fatal MSG error) raising an unclear exception. It now uses InvalidFileFormatError.
• Updated README to contain documentation on command line option added in previous version.
• Removed MSGFile.mainProperties after deprecating it in v0.36.0.
• Added new Properties Intelligence type: ERROR. This type is used when a properties instance is created but has something wrong with it that is not necessarily fatal. Currently the only thing that will cause it is the properties stream being 0 bytes.

v0.36.5

• Documented and exposed to the command line the ability to save the headers to it's own file when saving the msg file. Thanks to martin-mueller-cemas on GitHub for fixing this (and telling me I can switch the branch a pull request points to). I don't know when I added the code to allow the user to do it, but somehow it was never documented and never exposed to the command line.

v0.36.4

• [TeamMsgExtractor #291] Fixed typo in MSGFile.saveRaw that may have existed for a significant amount of time. It was using the wrong function (same name, but with different capitalization) but was hidden until MSGFile stopped being derived from OleFileIO.
• Added logging code to MessageBase.getSavePdfBody to log the list that is going to be used to run wkhtmltopdf. This is mainly for debugging purposes, to allow users to potentially see why their arguments may be failing.
• Updating funding information on GitHub and the README with more ways to support the module's development.
• Fixed one of the exceptions in MessageBase.getSavePdfBody not using an fstring which caused it to omit information.
• Changed the way wkhtmltopdf is called to patch a possible security vulnerability. This also seems to have fixed [TeamMsgExtractor #291].

v0.36.3

• Added an option to skip the body if it could not be found, rather than throwing an error. This will cause no file to be made for it in the event no valid body exists. For the save functions, this option is skipBodyNotFound and from the command line the option is --skip-body-not-found.
• Fixed a bug that caused contacts to save the business phone with two colons instead of 1.

v0.36.2

• [TeamMsgExtractor #286] Fix missing import.

v0.36.1

• [TeamMsgExtractor #283] Added file for typing recognition.
• Added new option to allow messages with NotImplementedError attachments to at least save everything not related to them. From the command line this would be --skip-not-implemented or --skip-ni. From the save function, this would be the skipNotImplemented option.

v0.36.0

• Improved type hints to tell when a function may not return anything.
• Added support for the reportTag property to MessageBase. I noticed this was one of the properties for IPM.Outlook.Recall so I decided to implement it. I'll work on ensuring all of [MS-OXOMSG] and [MS-OXCMSG] are implemented at a later date, including splitting off REPORT into it's own class, because it is it's own class.
• Fixed a few minor issues in some properties. Mostly some bool returning properties should have been returning False when they were not found instead of None. Ones that may return None are specifically typed as optional in the source code. Unfortunately using the help command doesn't seem to show the return type for properties for me at least on Python 3.9 and below.
• Fixed Attachment.save returning a pathlib.Path object instead of a str after the conversion to pathlib.
• Added code to allow pathlib objects in utils.openMsgBulk. It uses glob.glob which cannot take a pathlib object.
• Removed PermanentEntryID from EntryID.autoCreate. It shares a provider ID with AddressBookEntryID, and as such would never generate from it anyways. If you specifically need a PermanentEntryID you will have to instantiate it manually. Additionally, the entry has been removed from enums.EntryIDType.
• Fixed the GUIDs for some of the EntryID structures being returned as bytes instead of a standard GUID string. This is considered a breaking change and is the reason for the full version increase.
• Improved consistency of properties (the Python kind, not the MSG kind) so that properties for the raw data used to create an instance will all use the same name. Not all classes have this, but the ones that do will now all use rawData as the property name.
• Fixed the properties header being read in a way that actually swapped the values in it.
• Modified MSGFile to use the same name for Properties instances as all other classes. MSGFile.mainProperties will currently raise a DeprecationWarning instead of outright failing to help ease the transition. Use MSGFile.props instead.

v0.35.3

• [TeamMsgExtractor #280] Fix typing issue in message_base.py.

v0.35.2

• Made a change to the argument handling for --no-folders. Since it requires --attachmentsOnly to work, I simply made it error when it's not given to avoid confusion.
• Updated README.
• Added -v to be used instead of --verbose. This allows you to do -vvv for verbosity level 3.

v0.35.1

• Fixed a few property conflicts that I missed in the last release (forgot to run the helper script before releasing).

v0.35.0

• [TeamMsgExtractor #206] Implemented full support for Post objects, including the ability to save them.
• [TeamMsgExtractor #212] Implemented full support for Task objects, including the ability to save them. This also includes TaskRequest objects.
• [TeamMsgExtractor #110] Implemented full support for Contact objects, including the ability to save them.
• [TeamMsgExtractor #143] Rewrote the system used for Appointment objects to include all of the objects specified in [MS-OXOCAL]. Name changed to AppointmentMeeting. Completed support for Appointment objects, including the ability to save them.
• [TeamMsgExtractor #243] Added optional dependency mimetype-magic (installable using the mime extra) which helps to identify attachments that do not give a mime-type.
• [TeamMsgExtractor #274] Apparently the properties stream can have random garbage at the end of it (outlook generate the file that showed this) so code was added to ensure it wouldn't break everything.
• [TeamMsgExtractor #274] Made sure that the save function would report if it failed to find or generate a valid body. Specifically, if you were just trying to use the plain text body but it didn't exist (the stream didn't exist, not that the stream was empty) it would silently pass, which was bad behavior. Additionally, allowFallback will change the message to specify that current options were not usable for getting a valid body.
• [TeamMsgExtractor #207] Changed behavior for max date. Apparently it looks like it is supposed to be August 31, 4500 at 11:59 PM. However, in case this needs to change, we have created a constant called extract_msg.constants.NULL_DATE to represent this that you can use in your code to not have to worry about changing your code if we check it.
• Moved a few more minor constants to enums.
• Added support for many internal data structures, specifically Entry ID structures.
• Refactored classes from extract_msg.data to submodule extract_msg.structures.
• Added python_requires to setup.py as I noticed that it was missing.
• Due to new saving requirements, adjusted the way header injection worked all around. Functions are now built-in to MessageBase. getSaveXBody functions have also been moved down to be defined in MessageBase. If the extension class needs to specify custom behaviors for creating the save bodies, these functions will need to be overridden.
  • For saving, MessageBase (being the lowest one to currently contain bodies) has a few new properties. These properties represent the injection strings that will be injected into the bodies for the header, with an additional property to specify what properties map to what part of the format string. See MessageBase.headerFormatProperties for more information and an example of how to implement this in your own class.
  • Injection strings in constants have been removed in favor of dynamic generation, which only creates what is needed. No you will no longer see an empty Bcc field in your messages when you save them.
  • Plain text bodies now also use this injection, making it easy to change the header in all bodies by overwriting a single property that tells the program what data to put where.
• Fixed issue in encapsulated RTF header that caused the "To" field to not be present. I had to write them by hand, so it was bound to happen. They are now dynamically generated by each instance, so these fields should always appear.
• All save code has been moved down from Message into MessageBase for convenience. Message exists now for specific checking and for future specializations. This also means that anything that is a MessageBase now has the entire framework for saving built-in, with easy way to change details.
• Fixed bad property in Contact.
• Created save function for Contact. Saving, though it exists, is rather minimal and is limited to plain text and HTML.
• Significantly extended the Contact class's properties.
• Adjusted the naming of a few Contact properties to better match the microsoft names.
  • firstName -> givenName.
  • lastName -> surname.
  • businessPhone -> businessTelephoneNumber
  • Etc.
• Changed existing fax properties to give a dictionary of the properties they actually contain rather than just the number. This makes them behave like the newly added email properties.
• Fixed issues with Task properties being incorrect.
• Added implementation for PtypErrorCode.
• Changed behavior of Properties.date to only return the submit time. This is to ensure messages that were never sent do not have a sent date.
• Changed behavior of MessageBase.date to only return a send date if the message has been sent. For messages with no flags, it assumes True.
• Generally brought saving behavior closer to the way outlook handles it.
• Made SignedAttachment and BaseAttachment more similar by adding properties to each that are shared. BaseAttachment now have a name property and SignedAttachment now have longFilename and shortFilename.
• Fixed issue in HTML saving that would cause some characters to be dropped when rendering them due to how the header injection worked.
• Removed __init__ methods from MSG classes that don't change it. This ensures notes are easily passed down.
• Correction to last comment, one max date was supposed to be at that date, but another max date is at a different date of the same year.
• Changed the way that PtypTime is handled, making it a single function in utils.
• Upgraded dependencies to newer versions (some really need to be newer, like tzlocal, for best results). Included dependencies are beautifulsoup4 and tzlocal.
• Fixed an issue where zip file naming conflicts always failed in zip files for attachments. Both the embedded msg and the plain attachments would fail.
• Actually fixed the issue that would break the main loop.
• MSGFile no longer inherits directly from OleFileIO. While I would prefer to do that, the __init__ method for it is rather expensive, and allowing embedded msg files to directly share each other's instances of OleFileIO would improve speed immensely.
• Fixed attachments not being preemptively loaded when delayAttachments was False.
• utils.openMsg now delays attachments while loading the file to get the class type. This means all time for attachments is cut in half as they are only ever loaded once. It also means that files that won't open due to attachments will error a little later, but this shouldn't be a problem.
• Changed named properties Guid back to constants. This has to do with the next entry.
• Fixed a major issue in named properties. Apparently the ID is not enough and you must have the GUID as well, as multiple properties can share the exact same ID.
• Added option --no-folders to the command line allowing you to save all attachments from a set of MSG files into a single folder.
• Added option --skip-embedded to the command line to skip saving embedded MSG files.
• Added option skipEmbedded to Attachment.save (and all other related save methods that call it) to skip saving an embedded MSG file.
• Changed __main__ so that it opens the zip file there instead of relying on everything it calls to do it again and again.
• Changed the behavior of --verbose to allow it to be stacked for more verboseness. Specifying it once turns on warnings, twice for info, and three times for debug. Not specifying it only turns on error logging.

v0.34.3

• Fixed issue that may have caused other olefile types to raise the wrong type of error when passed to openMsg.
• Fixed issues with changelog format.
• Fixed issue that caused progress to sometimes break the main loop when a file had Unicode characters if the console it was writing to didn't support them.
• Added option to MessageBase (and subsequently openMsg) that allows you to override the code being used for deencapsulation. See MessageBase.__init__ for details on how to create an override function.

0.34.2

• [TeamMsgExtractor #267] Fixed issue that caused signed messages that were .eml files to have their data field not be a bytes instance. This field will now always be bytes. If a problem making it bytes occurs, an exception will be raised to give you brief details.
• Added function utils.unwrapMultipart that takes a multipart message and acquires a plain text body, an HTML body, and a list of attachments from it. These attachments are returned as dicts that can easily be converted to SignedAttachments. It replaces the logic mailbits was being used for, and as such mailbits is no longer required. The module may be reintroduced in the future.
• Added new property emailMessage to SignedAttachment, which returns the email Message instance used to get data for the attachment.