Dry run
- Standard term
A dry run is a practice mode where a job goes through its steps and reports what it would change, and why, while leaving your data as it was.
The job picks targets by its usual rules, then skips each delete, change, or send and continues, so a mistaken rule shows up on the list before anything is gone. The skip is safest written into the tool's code, since an agent only told to hold back can still delete by mistake.
Where you’ll meet it
Your coding agent has written a script to clear duplicate photos out of your library, about 3,000 by its count. Some of those “duplicates” are edited versions you made of the same shot, and you're not sure the script treats them differently from true copies.
What to do
Have the agent add a dry-run setting to the script that skips the delete step and lists each photo it would remove beside the one it takes for the original, then show you that list. If your edited shots are on it, have the agent fix the matching rule and dry-run again.
Easy to mix up
| Dry run | Sandbox |
|---|---|
| On your actual files, as they are now | In a walled-off space, away from your files |
Check yourself
A dry run lists 200 old files to delete and reports no errors. Nothing changes before the real run, which stops halfway with “permission denied.” Which explanation is likelier?
Go deeperDry runs differ by tool, and their lists go stale
Programs run by typed commands offer the mode under names like --dry-run, and tools that set up servers, the computers that host sites and apps, show it as a plan you approve first. A dry run may still read data, call services that only check, or write log files, so read what a tool's mode does before trusting it with live data.
The plan describes the targets as they were when it ran. Files and records can change before the real run, so careful tools save the approved plan and have the real run apply exactly that list, or stop when its targets no longer match.