Day13 - Implicit Knowledge

In today's post I want to pay attention to a topic I stumbled across last week. As you probably know, I have extended Ugurcan's trello-board visualization app for Gitub-Projects support. In order to visualize the tasks/issues from a Github board, I use the Github REST API to fetch all issues from a given repository. At least I thought I'd fetch all issues...

At first, everything seemed fine, but two days later, I noticed that the visualization tool must be broken somehow. The reason for this was that Github limits the number of issues in the HTTP response to 30 items by default. This explains why the app worked as expected only in the first days (when my board had less than 30 issues). In order to retrieve all issues of a repository, the Github-API provides Pagination via a HTTP header called Link. Since manual pagination is a bit complicated and error-prone, I decided to integrate a JavaScript library which already implements pagination. So, everything is fine now?

I don't think so! First of all, it may not be obvious to other developers why I am using an external library, as I only use a minimum of the its functionality. What happens if another developer decides to remove this Github client libary or replaces it with another one which maybe has not support pagination? The error would only be noticed if a repository with more than 30 issues was tested again. So the question is how can we share implicit knowledge to others?

For this I came up with the following possibilities:

1. Write additional code (in this case the pagination part) on my own

This makes it 100% clear to others that Github uses pagination, so it gets explicit. However, this leads to more code, which again is bad with regard to maintainability or errors proneness.

2. Add a comment in the code which explains why we use an external lib for the API call

I'm not a fan of comments, maybe becase good comments are so hard to write. Unlike code, comments are often not maintained. Comments can simply be lies or contain misleading information which may confuse the reader.

3. Add a software test which uses a dummy Github repository which contains more than 30 issues This could be too much effort? What would happen when Github decides to change their default limit of 30 items? In addition, software tests with external systems could fail for a lot of other reasons (no internet connection, server not reachable etc).

4. Build a wrapper-function around the library call and give it an appropriate name

I decided to wrap the function call of the library in an additional function called 'fetchAllIssues'. This way I hope (and this hope feeling tells me, that this cannot be the silver bullet!) that other developers realize that the library returns ALL issues, not just some.

5. State the problem in the commit message.

I've also done that, but I think no one ever will read it.

6. Ostrich method

Simply stick your head in the sand and pretend there is no problem... This is one of the worst options I think!

Do you know a better solution how to make implicit things in code more explicit? I think it would be great to hear some suggestions on this topic!

Cheers!

Markus