I had this stuck in my head for months, but only now managed to have some time to write it up
Below are two snippets of SQL code that do the same thing, namely extract the number of contact emails per year from a CRM. The result in both cases is something like the following:
+----------+------+ | count(*) | y | +----------+------+ | 1142 | 2012| | 1148 | 2013| +----------+------+
The first query is written against a heavily customized SugarCRM instance. The second is written against my own home-made CRM
SELECT count(*), year(email_addresses.date_created) y FROM email_addresses LEFT JOIN email_addr_bean_rel ON email_addr_bean_rel.email_address_id = email_addresses.id WHERE email_addresses.email_address LIKE '%@%' AND email_addr_bean_rel.bean_module = 'Contacts' AND email_addr_bean_rel.deleted = 0 AND email_addresses.invalid_email = 0 AND email_addresses.opt_out = 0 AND email_addresses.deleted = 0 GROUP BY y;
SELECT COUNT(*), YEAR(created_at) AS y FROM clients WHERE email IS NOT NULL GROUP by YEAR(created_at);
It should be obvious even for a non-programmer which is easier to understand. What is not obvious for non programmers is the amount of effort one needs to make before SugarCRM starts doing things it was not designed for. I could now start talking about the importance of data structures and how the UI is irrelevant, but this will take more pages than Google can handle.