| Author: | Stefano |
|---|---|
| category: | Python, Web frameworks |
I wanted to use Grok for my future development, so I just spent a week trying to do porting of a project from Django to Grok. I had to back off.
Grok is very powerful without doubt. However, the learning curve is steep, and I am not talking about "mountain trekking" steep, I'm talking about "walking on the ceiling" steep. I was able to port basic functionalities, but I felt less and less confident about future development needs. With some effort, I could have learned how things work, but presenting such hurdles to any other future collaborator would have made things more complex both for me and them.
What I find most difficult is the logic. Innocuous python refactoring can have disruptive results (e.g. nothing works anymore). The error messages are difficult to understand because in order to understand them you have to know how things work. You have a lot of magic places and formulas you have to know and adhere to, and things are easy until you stay within the limits of this magic. If you want to leave this path, even just by a little, you are faced with the need to study additional libraries you didn't even know they existed, and in some case are not under the grok framework package. Interrelationships are tight, but it's difficult to understand how things interact. You cannot "evolve" your understanding step by step, and until you don't get a large part of the component network, you will probably be reluctant to even rename a class or a method. It's like walking on a minefield, with the difference that the minefield is n-dimensional: you do have a lot of power, but that comes with a huge cost in manhours attached. Lack of a bird's eye view of the framework and an intermediate step between the tutorial and the developer's documentation is another big issue. Not many examples can be found on the internet, and the ones that are available, are either too simple, or too complex, hardly explaining the reason behind particular lines of code.
Please don't interpret this post as a rant against Grok. Grok is not bad at all, it's simply big. As a consequence, it's difficult to handle for a single developer, unless he spent a fair amount of time studying it, knowing all its magic, when this magic works, and when it doesn't.This goes against one of my needs for the project: easy to understand, learn and modify by third parties.
I would like to thank the people at #grok on freenode IRC, in particular trollfot (author of the dolmen project, as far as I got) and goschtl. They were very helpful and friendly.