WTF – Scheduled Reports
Working on scheduled reporting systems, you expect to see the same concepts. What format wiil we send the report in? How will it be delivered? I opened the “scheduled report” table to see how this particular system had been implemented. I couldn’t immediately find which column i was looking for. In fact the only one that looked like it might be useful was something called deliv_method.
deliv_method
206
774
206
206
Ok, so we’ve got a numeric list of delivery methods. Nothing too unusual about that. I just need to find the list either in the database or codebase. But where do we store the format of the report? I worked my way through the report scheduler code until i found this.
$delivMeth = ($row[‘DELIV_METHOD’] >> 8);
$myFormat = ($row[‘DELIV_METHOD’] & 0xFF);
Are those bitwise operators? And even a literal hexadecimal number? The only place i’ve ever seen a bitwise operator is when setting the error reporting level. I’ve looked at this a few times and am still not sure i correctly understand it.
The answer eventually revealed itself.
$a = array(
0×0104 => "Email HTML",
0×0204 => "Email HTML Attachment",
0×0284 => "Email HTML Zipped attach",
0×0201 => "Email PDF",
0×0203 => "Email TSV delim",
0×0283 => "Email TSV delim zipped",
0×0203 => "Email CSV delim",
0×0283 => "Email CSV delim zipped",
0×0304 => "FTP HTML",
0×0301 => "FTP PDF",
0×0303 => "FTP TSV delim",
0×0303 => "FTP CSV delim",
);
Pretty clever you have to admit! Unfortunately it can lead to same rather ugly code.
if ((($('#INPUT\\[DELIVERY_METHOD\\]').val() & 0xFF00) >> 8) != 0×03){
Glue code
Glue is what holds your application together. Code needs to talk to databases, Javascript is calling your PHP functions using Ajax. When two systems, or different parts of the same system need to talk to each other, you end up with glue. It (hopefully) doesn't have any business logic. It's used or copied by at least a few different parts of your system, perhaps even in different projects in the same organisation.
Glue code holds things together. The only problem is that glue is not your application. Unless you're developing a framework, operating system or perhaps something that needs plugins, glue is not your strength. There are always exceptions but if you are writing glue, especially the same glue for the 3rd or 4th time, then you're not directly working on the application you are building.
Having said that, your project won't survive without glue. So how do you keep everything together?
- Recognise glue. The first step is to understand that you are now writing glue code and not working on business requirements. For example, say you are writing a script to process files that are uploaded to your server. Discovering the files is glue, the processing of then is specific to your application.
- Standardise across your organisation. Glue may start out as a one off but as time moves on it will become apparent which pieces are worth time improving and which aren't.
- Make it as simple to use as possible. It will need to be understood by a larger group of developers than your other code. Put that extra time into documentation and code beautification.
- Make it as robust as possible. You don't want this getting in the way of your 'actual' work.
- Use 3rd party glue. Someone has nearly always been before you. Wiring your PHP and JS together with AJAX? It's been done. Creating a DB access class? Been there, done that. If you think you can do better your probably wrong. You may write better code, but you probably won't document or test it as thoroughly as an open source project has already been.
In conclusion, you want to keep glue simple and straight forward. Recognise what is and isn't glue in your application. Share the glue you have with every one in your team, and if your brave, the outside world.
The idea of glue is described in Eric Raymond's excellent book The art of unix programming.