Re: [Patch v5 2/4] emacs: postpone a message

[Mark Walters <markwalters1009@gmail.com>]

On Sat, 12 Nov 2016, David Bremner <david@tethera.net> wrote:
> David Bremner <david@tethera.net> writes:
>
>> From: Mark Walters <markwalters1009@gmail.com>
>
> This really Mark's work, that I have split out into a separate file.
>
>> +(defcustom notmuch-draft-tags '("+draft")
>> +  "List of tags changes to apply to a draft message when it is saved in the database.
>
> Here and a few other places the documentation uses "the database" to
> mean the directory hierachy containing mail messages + the xapian
> database.  At some point I would like to be able to distinguish between
> the database and the maildir root (which doesn't need to be maildirs)
> when talking about configuration. I don't really know better terminology
> here, but I thought I would mention it in case someone else is inspired.

I think we could use "mailstore" for the maildir root, "database" makes
sense for the xapian database. However, I don't have a good term for the
combined whole -- so I am  not sure what would make sense in this
particular case.

>
>> +(defun notmuch-draft--mark-deleted ()
>
> This -- naming convention is my contribution. Perhaps eventually we
> could mark all private functions not intended to be called by users this
> way. Since it is essentially cosmetic, I didn't want to do that
> now. Indeed, one could quibble about the correctness of calling a
> function private and then putting it in a public hook.

I like it, and I agree. I do intend to remove it from the public hook
later (see below).

>> +(defun notmuch-draft-quote-some-mml ()
>> +  "Quote the mml tags in `notmuch-draft-quoted-tags`."
>> +  (save-excursion
>> +    ;; First we deal with any secure tag separately.
>> +    (message-goto-body)
>> +    (when (looking-at "<#secure[^\n]*>\n")
>> +      (let ((secure-tag (match-string 0)))
>> +	(delete-region (match-beginning 0) (match-end 0))
>> +	(message-add-header (concat "X-Notmuch-Emacs-Secure: " secure-tag))))
>> +    ;; This is copied from mml-quote-region but only quotes the
>> +    ;; specified tags.
>> +    (when notmuch-draft-quoted-tags
>> +      (let ((re (concat "<#!*/?\\("
>> +			(mapconcat 'identity notmuch-draft-quoted-tags "\\|")
>> +			"\\)")))
> One "hidden feature" is that regex characters in the quoted tags will be
> interpreted. Possibly calling regexp-quote instead of identity would be
> extra cautious here?

Good catch: I have made it regexp-quote as you suggest.

>> +(defun notmuch-draft-save ()
>> +  "Save the current draft message in the notmuch database.
>> +
>> +This saves the current message in the database with tags
>> +`notmuch-draft-tags` (in addition to any default tags
>> +applied to newly inserted messages)."
>> +  (interactive)
>> +  (let (;; We need the message id as we need it for tagging. Note
>> +	;; message-make-message-id gives the id inside a "<" ">" pair,
>> +	;; but notmuch doesn't want that form, so remove them.
>> +	(id (concat "draft-" (substring (message-make-message-id) 1
>> -1))))
>
> what do you think of isolating this code and commentary in a private
> function?

I have done this.

>> +     (if (member 'Message-ID message-deletable-headers)
>> +	 (progn
>> +	   (message-remove-header "Message-ID")
>> +	   (message-add-header (concat "Message-ID: <" id ">")))
>> +       (message "You have customized emacs so Message-ID is not a deletable header, so not changing it")
>> +       (setq id nil))
>
> I'm not sure if it's just me, but I find the (if (progn ...)
> else-clauses) a bit off-putting. An alternative would be to use cond
>      
> (cond
>  ((member 'Message-ID message-deletable-headers)
>   (message-remove-header "Message-id")
>   (message-add-header (concat "Message-ID: <" id ">")))
>  (t
>   (message "You have customized emacs so Message-ID is not a deletable header, so not changing it")
>   (setq id nil)))

I am happy either way, so I have made the change

>> +(add-hook 'message-send-hook 'notmuch-draft--mark-deleted)
>
> Can we avoid this by adding some code notmuch-mua-send-common?

Yes. We should do the same for notmuch-message-mark-replied. However
getting it right looks slightly non-trivial. In particular, the user can
abort sending for various reasons (eg charset things, or they have
message-confirm-send set) and we should not do the marking if they do
abort. However, message-send-and-exit will have killed the buffer so we
need to store the relevant local variables before we run that command.

Having said that, I think that message-send-hook is run before the send
anyway, so putting both of the above in notmuch-mua-send-common would
not make the situation any worse.

My inclination was to fix it properly afterwards -- but we could do the
above now as a "no worse than now" change.

Best wishes

Mark






>
>> --- /dev/null
>> +++ b/test/T630-emacs-draft.sh
>> @@ -0,0 +1,42 @@
>> +#!/usr/bin/env bash
>> +test_description="Emacs Draft Handling"
>> +. ./test-lib.sh || exit 1
>
> Ok, now I remember I wrote this part, so someone else should review.
> _______________________________________________
> notmuch mailing list
> notmuch@notmuchmail.org
> https://notmuchmail.org/mailman/listinfo/notmuch