First of all, you should definitely switch to using the Page Object pattern and keep your page objects under a separate directory - I think it's recommended to call the directory po.
Here is a sample for you, the project structure we currently have:
$ cd e2e
$ tree -L 1
.
├── config
├── db
├── helpers
├── mocks
├── po
└── specs
config is special directory where we keep our protractor configs - there could be multiple configs - for instance, for local testing and for testing on, say, BrowserStack.
helpers is, basically, our "libs"/"utils" directory. We keep custom jasmine matchers, additional "helper" modules with helper functions. Also, we have localStorage and sessionStorage modules that are convenient wrappers around window.localStorage and window.sessionStorage objects.
mocks is a directory where we keep protractor-http-mock mocks.
po is a directory where Page Objects are defined. Each Page Object in a separate file.
specs is where all of our specs live - they are organized into subdirectories logically.
Some of the helpers libraries are made globally available via global, example:
onPrepare: function () {
global.helpers = require("../helpers/helpers.js");
// ...
},
Also, to make the helpers and po import more convenient and avoid traversing the directories up in the tree and to better handle the nestedness, we've switched to using requirePO and requireHelper helper function suggested by @Michael Radionov, see:
I also really like the idea, proposed by @finspin, to make a node package out of each Page Object.
