Migrating to v3.x

Git-Tool v3.x is a major release which makes some fundamental changes to the way that we configure services for Git-Tool and how we choose where to place them on the filesystem. This guide will walk you through the process of updating your existing Git-Tool configuration to work with Git-Tool v3.x.

This document shows an example of an updated configuration, and then describes the specific changes you should expect to make when updating to v3.x.

For more information on why we've made these changes, take a look at the why we're changing things section.

Example

Here is an example configuration file which has been updated to reflect the changes introduced in Git-Tool v3.x.

---
directory: C:\\dev
services:
-  - domain: github.com
-    website: "https://{{ .Service.Domain }}/{{ .Repo.FullName }}"
-    gitUrl: "[email protected]{{ .Service.Domain }}:{{ .Repo.FullName }}.git"
-    httpUrl: "https://{{ .Service.Domain }}/{{ .Repo.FullName }}.git"
+  - name: gh
+    website: "https://github.com/{{ .Repo.FullName }}"
+    gitUrl: "[email protected]:{{ .Repo.FullName }}.git"
     pattern: "*/*"
+    api:
+      kind: GitHub/v3
+      url: https://api.github.com

-  - domain: dev.azure.com
-    website: "https://{{ .Service.Domain }}/{{ .Repo.Namespace }}/_git/{{ .Repo.Name }}"
-    gitUrl: "[email protected]{{ .Service.Domain }}:v3/{{ .Repo.FullName }}.git"
-    httpUrl: "https://{{ .Service.Domain }}/{{ .Repo.Namespace }}/_git/{{ .Repo.Name }}.git"
+  - name: ado
+    website: "https://dev.azure.com/{{ .Repo.Namespace }}/_git/{{ .Repo.Name }}"
+    gitUrl: "[email protected]:v3/{{ .Repo.FullName }}.git"
     pattern: "*/*/*"
apps:
  - name: shell
    command: sh
    default: true
  - name: code
    command: code.cmd
    args:
      - .
  - name: make
    command: make
    args:
      - build
    environment:
      - CI_SERVER=0
      - REPO={{ .Repo.FullName }}
aliases:
-  gt: github.com/SierraSoftworks/git-tool
+  gt: gh:SierraSoftworks/git-tool

Changes

Configuration Schema

Git-Tool's configuration schema has been updated from v1 to v2. There are several key changes, including the following:

  1. The domain field has been renamed to name.
  2. The httpUrl field has been removed in favour of using the gitUrl field to determine whether HTTP transport is used.
  3. The api field has been added to enable the creation of remote repositories on GitHub Enterprise servers.
  4. The removal of the http_transport feature flag, which is no longer supported.
# yaml-language-server: $schema=https://schemas.sierrasoftworks.com/git-tool/v2/config.schema.json
services:
    # The new layout for a service entry in your configuration file
  - name: gh
    website: "https://github.com/{{ .Repo.FullName }}"
    gitUrl: "[email protected]:{{ .Repo.FullName }}.git"
    pattern: "*/*"
    api:
      kind: GitHub/v3
      url: https://api.github.com

    # The original layout for a service entry in your configuration file
  - domain: github.com
    website: "https://{{ .Service.Domain }}/{{ .Repo.FullName }}"
    gitUrl: "[email protected]{{ .Service.Domain }}:{{ .Repo.FullName }}.git"
    httpUrl: "https://{{ .Service.Domain }}/{{ .Repo.FullName }}.git"
    pattern: "*/*"


 
 
 
 
 
 
 
 







Specifying Services

Git-Tool v3.x changes the way you specify a service as part of a repository's name.

# v2.x
- github.com/sierrasoftworks/git-tool
# v3.x
+ gh:sierrasoftworks/git-tool

This affects the way you use commands like gt open and gt new and will require that you update your alias definitions to match.

  aliases:
-   gt: github.com/sierrasoftworks/git-tool
+   gt: gh:sierrasoftworks/git-tool

Folder Structure

As a side-effect of changing the naming scheme used by services in Git-Tool, you may need to move your repositories around on disk. This is only required if you wish to switch to the new naming scheme (github.comgh).

To do so, simply rename the corresponding service directories inside your development directory to match the new service name field.

Why are we changing things?

Git-Tool, as with all software, is a constant work in progress and as we continue to use it we find new and fun ways to break it, or edge cases that we didn't consider when we first designed it. Over time, these have become the paper-cuts that have pushed people away from using it as their one-stop solution for repository management.

We're hoping that by making these changes, we alleviate many of these pain points and make Git-Tool a more useful tool for everyone. Some of the key issues that kept cropping up included:

  1. Anonymous Pulls from GitHub

    GitHub's public repositories play really nicely with anonymous HTTPS cloning, however they tend to get uppity when you use an SSH key if you aren't one of the maintainers. A quick fix for this was to switch on the http_transport feature flag temporarily, but we often don't want to rely on that for our private repositories.

    Being in a position to easily choose which transport we use for a given repo would massively simplify this, but using domain as an identifier for a service and separating the gitUrl and httpUrl fields made doing so quite the hack.

    To make this a bit cleaner, we opted to rename and simplify these fields to make that workaround the primary, supported, way of running things.

  2. Supporting GitHub Enterprise

    One of Git-Tool's coolest features is the way that gt new integrates with GitHub, automatically creating a new repository whenever you run the command. Unfortunately, coupling this functionality to the domain of the service in your config really wasn't the most flexible way to handle things and meant that even though GitHub Enterprise uses the same API, we didn't include support for it.

    By separating the notion of "which API should I use" from the concept of which domain we connect to, we can make supporting GitHub Enterprise much easier and unlock the door for similar integrations on other self-hosted Git platforms.

  3. Long File Paths

    Git-Tool's default folder structure is great when it comes to being able to navigate and find what you're looking for, but it is rather verbose. This was a particularly large problem for Windows users who didn't have Long Path Support enabled (limiting them to 256 characters).

    By shortening the ID used to identify a service, we can save a few extra characters and hopefully avoid making their lives too hard.

  4. Name Resolution Conflicts

    With us allowing arbitrary names for services, it was likely that a user would add a service called bob and then try to add a repository called bob/repo to their GitHub service. We'd now struggle to determine whether they were asking for the gh:bob/repo or the bob:repo repository. To prevent this, we've changed the separator between service and repository to be : instead of the original /, making it clear which component we're referring to.