Notmuch 0.18 (2014-05-06)


This new release includes some enhancements to searching for messages by filesystem location (folder: and path: prefixes under General below). Saved searches in Emacs have also been enhanced to allow distinct search orders for each one. Another enhancement to the Emacs interface is that replies to encrypted messages are now encrypted, reducing the risk of unintentional information disclosure. The default dump output format has changed to the more robust batch-tag format. The previously deprecated parsing of single message mboxes has been removed. For detailed release notes, see below.


The folder: search prefix now requires an exact match

The folder: prefix has been changed to search for email messages by the exact, case sensitive maildir or MH folder name. Wildcard matching (folder:foo*) is no longer supported. The new behaviour allows for more accurate mail folder based searches, makes it possible to search for messages in the top-level folder, and should lead to less surprising results than the old behaviour. Users are advised to see the notmuch-search-terms manual page for details, and review how the change affects their existing folder: searches.

There is a new path: search prefix

The new path: search prefix complements the folder: prefix. The path: prefix searches for email messages that are in particular directories within the mail store, optionally recursively using a special syntax. See the notmuch-search-terms manual page for details.

Notmuch database upgrade due to folder: and path: changes

The above mentioned changes to the folder: prefix and the addition of path: prefix require a Notmuch database upgrade. This will be done automatically, without prompting on the next time notmuch new is run after the upgrade. The upgrade is not reversible, and the upgraded database will not be readable by older versions of Notmuch. As a safeguard, a database dump will be created in the .notmuch directory before upgrading.

Library changes

Notmuch database upgrade

The libnotmuch consumers are reminded to handle database upgrades properly, either by relying on running notmuch new, or checking notmuch_database_needs_upgrade() and calling notmuch_database_upgrade() as necessary. This has always been the case, but in practise there have been no database upgrades in any released version of Notmuch before now.

Support for indexing mbox files has been dropped

There has never been proper support for mbox files containing multiple messages, and the support for single-message mbox files has been deprecated since Notmuch 0.15. The support has now been dropped, and all mbox files will be rejected during indexing.

Message header parsing changes

Notmuch previously had an internal parser for message headers. The parser has now been dropped in favour of letting GMime parse both the headers and the message MIME structure at the same pass. This is mostly an internal change, but the GMime parser is stricter in its interpretation of the headers. This may result in messages with slightly malformed message headers being now rejected.

Command-Line Interface

notmuch dump now defaults to batch-tag format

The old format is still available with --format=sup.

notmuch new has a --quiet option

This option suppresses the progress and summary reports.

notmuch insert respects maildir.synchronize_flags config option

Do not synchronize tags to maildir flags in notmuch insert if the user does not want it.

The commands set consistent exit status codes on failures

The cli commands now consistently set exit status of 1 on failures, except where explicitly otherwise noted. The notable expections are the status codes for format version mismatches for commands that support formatted output.

Bug fix for checking configured new.tags for invalid tags

notmuch new and notmuch insert now check the user configured new.tags for invalid tags, and refuse to apply them, similar to notmuch tag. Invalid tags are currently the empty string and tags starting with -.

Emacs Interface

Init file

If the file pointed by new variable notmuch-init-file (typically ~/.emacs.d/notmuch-config.el) exists, it is loaded at the end of notmuch.el. Users can put their personal notmuch emacs lisp based configuration/customization items there instead of filling ~/.emacs with these.

Changed format for saved searches

The format for notmuch-saved-searches has changed, but old style saved searches are still supported. The new style means that a saved search can store the desired sort order for the search, and it can store a separate query to use for generating the count notmuch shows.

The variable is fully customizable and any configuration done through customize should just work, with the additional options mentioned above. For manual customization see the documentation for notmuch-saved-searches.

IMPORTANT: a new style notmuch-saved-searches variable will break previous versions of notmuch-emacs (even search will not work); to fix remove the customization for notmuch-saved-searches.

If you have a custom saved search sort function (not unsorted or alphabetical) then the sort function will need to be modified. Replacing (car saved-search) by (notmuch-saved-search-get saved-search :name) and (cdr saved-search) by (notmuch-saved-search-get saved-search :query) should be sufficient.

The keys of notmuch-tag-formats are now regexps

Previously, the keys were literal strings. Customized settings of notmuch-tag-formats will continue to work as before unless tags contain regexp special characters like . or *.

Changed tags are now shown in the buffer

Previously tag changes made in a buffer were shown immediately. In some cases (particularly automatic tag changes like marking read) this made it hard to see what had happened (e.g., whether the message had been unread).

The changes are now shown explicitly in the buffer: by default deleted tags are displayed with red strike-through and added tags are displayed underlined in green (inverse video is used for deleted tags if the terminal does not support strike-through).

The variables notmuch-tag-deleted-formats and notmuch-tag-added-formats, which have the same syntax as notmuch-tag-formats, allow this to be customized.

Setting notmuch-tag-deleted-formats to '((".*" nil)) and notmuch-tag-added-formats to '((".*" tag)) will give the old behavior of hiding deleted tags and showing added tags identically to tags already present.

Version variable

The new, build-time generated variable notmuch-emacs-version is used to distinguish between notmuch cli and notmuch emacs versions. The function notmuch-hello-versions (bound to 'v' in notmuch-hello window) prints both notmuch cli and notmuch emacs versions in case these differ from each other. This is especially useful when using notmuch remotely.

Ido-completing-read initialization in Emacs 23

ido-completing-read in Emacs 23 versions 1 through 3 freezes unless it is initialized. Defadvice-based Ido initialization is defined for these Emacs versions.

Bug fix for saved searches with newlines in them

Split lines confuse notmuch count --batch, so we remove embedded newlines before calling notmuch count.

Bug fixes for sender identities

Previously, Emacs would rewrite some sender identities in unexpected and undesirable ways. Now it will use identities exactly as configured in notmuch-identities.

Replies to encrypted messages will be encrypted by default

In the interest of maintaining confidentiality of communications, the Notmuch Emacs interface now automatically adds the mml tag to encrypt replies to encrypted messages. This should make it less likely to accidentally reply to encrypted messages in plain text.

Reply pushes mark before signature

We push mark and set point on reply so that the user can easily cut the quoted text. The mark is now pushed before the signature, if any, instead of end of buffer so the signature is preserved.

Message piping uses the originating buffer's working directory

notmuch-show-pipe-message now uses the originating buffer's current default directory instead of that of the *notmuch-pipe* buffer's.


nmbug adds a clone command for setting up the initial repository and uses @{upstream} instead of FETCH_HEAD to track upstream changes.

The @{upstream} change reduces ambiguity when fetching multiple branches, but requires existing users update their NMBGIT repository (usually ~/.nmbug) to distinguish between local and remote-tracking branches. The easiest way to do this is:

  1. If you have any purely local commits (i.e. they aren't in the nmbug repository on, push them to a remote repository. We'll restore them from the backup in step 4.
  2. Remove your NMBGIT repository (e.g. mv .nmbug .nmbug.bak).
  3. Use the new clone command to create a fresh clone:

    nmbug clone

  4. If you had local commits in step 1, add a remote for that repository and fetch them into the new repository.