Thank you, seems to be working as expected.

One thing I want to mention, I needed to add a mapping sent = "accountX/Sent" or else Himalaya tried to access a wrong directory again when saving a new mail to the database, but with the mapping it now works so that I can actually start using it.

The github url returns 404 for me. Permission issue?

If you have some ideas on a solution, I can also contribute to implement it. So far I haven't been really getting into the code of Himalaya, but I'd be happy to do so.

To save sent emails, in other clients what I've done is to set sendmail option to a script that first calls notmuch insert --folder=accountX/sent and then sends the mail using msmtp.

My notmuch config is very basic, the database option you mentioned is set to $HOME/mail and indeed there's $HOME/mail/.notmuch present. The folder structure is created by mbsync with Subfolders Verbatim setting, which I use to fetch the emails.

To query account1 inbox, I do notmuch search folder:account1/inbox.

Now that dates are parsed correctly, it seems maybe mailparse shouldn't be used for parsing headers either.

I have some (spam, don't ask why) email in my notmuch database that has a peculiar sender:

From: ""☰DHL☰" <censored@spammer>

Loading the email fails with:

Error: cannot parse notmuch sender ""☰DHL☰" <censored@spammer>

Caused by:
    Address string unexpectedly terminated

Notmuch backend seems to make an assumption that the database is in a maildir directory. I have a structure where my notmuch database is in a root folder, which then has bunch of different maildir folders under it.

mail .notmuch account1 inbox tmp cur new sent tmp cur new account2 ... ... account3 ... ...

Creating new emails with Himalaya doesn't seem to work with this kind of structure. An error "cannot store notmuch email" is thrown whenever trying to do operations that would write emails to disk.

In my notmuch database I have a message that has a datetime with timezone (CET) which isn't supported by dateparse by design1.

This prevents Himalaya from loading the database at all.

It looks like the bug has been fixed, but I find it's still worthwhile to have the zz called to actually see the result when evaluating the last line. In your screencast it looks like the line is evaluated while you type, which causes the issue. This is not something that happens for me.

Maybe we could have an option to disable this behaviour.

Discussed here