Team Workflow
The scenario
John builds the data model, Adam and Eve build sheets. All three use the same Git provider and organization in their Gitoqlok profiles.
John creates the master app and the repository
John builds the app, opens Gitoqlok and clicks Create Repository. Gitoqlok creates the repository with the main branch and John commits the script and the data model. He publishes the app to a development stream or shares it in a space.
Adam and Eve take copies
Each duplicates the published app into their own workspace. When they open their copy, Gitoqlok detects the master app and offers Create Branch. Adam's branch is adam, Eve's is eve (the pattern $qlikUserName names branches automatically).
They work and commit independently
Adam builds the pivot sheets, Eve the KPI sheets. Each commits to their own branch as often as they like. Nothing they do affects the master app or each other.
Pull requests
When a feature is done, the developer opens a pull request on GitHub, GitLab, Bitbucket or Azure DevOps from their branch into the main branch. The reviewer sees one JSON file per sheet or master item, so changes are readable. John merges.
John updates the master app
Back in the master app the popup shows that the repository is ahead. John clicks Update and the merged sheets appear in the app. He republishes with Publish App → Replace existing app.
Adam and Eve catch up
To get each other's work, they either duplicate the master app again (a fresh branch) or use Import to pull specific sheets from the main branch into their copies.
Rules that keep merges painless
- The main branch is the published app. Only the app owner (or the release manager) updates and publishes it.
- One naming convention for branches — pick one (
$qlikUserName,feature-…,ticket-123) and stick to it. - Divide the work by layer and by sheet. One developer on the load script at a time; two developers never edit the same sheet in parallel.
- Merge with pull requests, not by copying objects between apps. When a pull request conflicts, use the Diff Viewer and Load Script Diff to understand the difference and Import to take the other side's objects, then commit the resolved version and reopen the pull request.
- Link commits to tickets (
#123in the message) so the history explains why. - Do not delete branches. They are the history of the copies and may be needed for a roll back.
- Reduce noise in diffs. Turn on Sort elements on sheet to keep order, Do not commit full loadScript.qvs file and Use original Qlik Application Id in thumbnail URLs in Repository Settings.
- Organize the script into reusable sections and import them into other apps instead of copy-pasting.
Setting up the Git side
- Use an organization (GitHub organization, GitLab group, Bitbucket workspace, Azure DevOps project) rather than personal accounts, so every developer sees the repositories.
- Give every developer write access and the right to create branches. On GitLab, allow Developers to push to the default branch or route everything through pull requests.
- If a personal repository must be used, add the other developers as collaborators.
- Every developer's Gitoqlok profile must point to the same provider and organization.
- Bitbucket Cloud: turn off the 100 MB push limit in the repository settings if QVF files are committed.
