Often, we have a need to collect data from customers. Immediate thinking is to make an app change, push it up to store, get people to upgrade (doesn't happen!) and then pray. A little more savvy version of this is making a web app and then send the links with a uuid based customer identification.

Why not just ask users to email you the data? People know how to use email! Following snippet shows how easy it is to automate fetching the data and then sending to a script to process.


Two venerable unix utilities.

  • fetchmail supports a wide variety of protocols to get email from a remote email server.
  • procmail do something with your email. Including routing to specific mailboxes or commands.


If you are in recent ubuntu or its derivates, just install these two.

sudo apt-get install fetchmail fetchmailconf procmail


Let us create a fetchmail config file to get data from gmail.


First, Configure gmail to support imap.

To make things easier, you should also setup a gmail filter to move emails matching your criteria to a label, from inbox. This way, we can have multiple configs for multiple use cases all using same gmail account.

For example, I have one gmail account called myemail@mydomain.com. So, for usecase A, I ask people to email to myemail+a@mydomain.com_ and for usecase B, they send to myemail_+B@mydomain.com_. In gmail, you simply add filters based on To: field in the incoming mail.


fetchmail.rc is below.

poll imap.gmail.com protocol IMAP
   user "myemail@mydomain.com" is localuser here
   password 'whateveritis'
   folder 'mygmaillabel'
  • fetchall indicates we should get all unread mails from this folder.
  • keep ensures mails are left on the server and not deleted. You can use nokeep if you want to delete the mails we've downloaded.


procmail.rc is below.

! `/full/path/to/processemail.py`

What this is doing is that every email that comes to procmail will be passed to this script processemail.py via stdin. There are many samples out there - for example, download attachments using the cool walk method.

Most of the examples also have code for reading from gmail directly, but I prefer keeping getting data and processing data as separate things.

Note that the configuration above is completely reliant on the script - if your script has an error and threw out the content, then fetchmail would have marked the mail as read and you would've lost the content from local.


Just execute a command as below.

fetchmail -f fetchmail.rc \
  --mda "procmail ${PWD}/procmail.rc"

Straight forward, isn't it? Get the mail and send to mail delivery agent, which is procmail using your config.

Add this to a cron and you are done. fetchmail comes with its own daemon mode too if you prefer that route.

I like to cron things on my own since for different use cases, I have different config files. It is far easier that way than adding multiple filters in procmail to direct to different scripts.