openwith VS .emacs.d

Just kidding. Of course you can open files with a key binding. I personally use a hacked version of openwith package. I wanted exactly same thing as you, to be able to open a file in external application with a key shortcut, so I have hacked original openwith.el to achieve this. My setup is like this: I press S- to open a file in external application. If I press just it will open a file in Emacs itself. So S- on a pdf file opens it in Okular in gnu/linux and just opens a pdf in Emacs with pdf-tools. You can see my setup on github if you are interested. Scroll down to * Pacakges/Dired.

In my Helm, I have to actively choose the candidate to confirm it. So I can type in both paths that are shorter or longer then existing ones. I even made a video to demonstrate it, the thread was relatively recently up I think. My Helm setup is here it if helps you, find Helm in the list of packages.

I also use Helm, and I have no problems. Just keep typing, once you typed a letter that does not exist in a path name it will stop completing. I don't know if I have some special option enabled/disabled; I don't think you need it, but you can see my Helm config (just scroll down untill you find "Helm").

When all deps are installed,my config is over 200 packages. On my Arch Linux desktop I built in 2016, with i7 4.6k (haswell) it starts ~0.7 secs, but init time will be anything between 0.5 ~ 0.8 secs, i guess depending on what system does. So all things same, init time will vary.

I am using it in my literate org-config. If you scroll down, there is a big list of packages, and I have done a small wrapper around helm-imenu, to jump to a package configuration. Looks like this.

That does not work for completions, at least not for me. It works for keymaps, so you can have mode specific (or really any) keymap in src blocks. I have been using his method myself in my init file generator for quite a while now. If you (or anyone) knows/have an idea how to expand it for completions and eldoc, I would be really happy to hear.

I don't know what if it is more robust but I use more or less plain dired with just some options turned on to make it less noisy to look at, but I don't "manage" my files so much to be honest. I do use some extras from dired-hacks, and my own dired-auto-readme, but that is about it. You can check my setup if you wish, look at "dired" under packages and in Lisp folder for "dired-extras.el".

I personally put all custom lisp in a special directory and scrape autoloads myself. If you are curious, you can check under "generator", functions generate-autoloads and collect-autoloads, but there is nothing special, just plain text search and copy-paste programmatically. I don't recommend to use it though.