pattern playground

Dry run

Added · Seed

The actor has written something the system will apply to many cases: a mail rule, a find-and-replace across a project, a rename over a folder of files, a formula in a column, a template that will be merged with a list. Reading it does not say what it will do. The cases it will meet are either out of view or have not arrived yet, and a mistake lands on all of them at once. A dry run applies the thing the actor wrote to real or sample cases and shows what would happen, without letting it happen.

Replace-all with a preview of every occurrence, a batch rename that lists old and new names side by side, a mail-merge preview, a regex tester, an automation’s “test this step”, and replaying a mail rule against the inbox are all this move.

Forces

  • Representative vs. cheap — the cases that decide whether the rule is right are the ones near its edge, and a random or first-N sample rarely holds them. A country-music rule previewed against a library that is mostly not country shows no country at all.
  • Rehearsal vs. the real run — what the run shows is true of the cases it ran on. The rule meets other cases later, and the actor’s confidence from the rehearsal carries over to them.

Forms

  • Replay the past. Apply the thing written to what already happened and show what would have come out differently. This is the form shown to improve people’s success. It works because the cases are real and the actor already has opinions about them.
  • Run against the standing set. Step a new rule through alongside the rules already in place, so conflicts and loops show up before saving. This is the only form that reaches the interaction between one rule and the list it joins.
  • Show the edge cases. Show the cases that now match for the first time, the cases that match but only just, and the cases that miss but only just. This is the form to use when there is no history to replay.
  • Live result. When the thing written applies to a visible collection, the result updates as each part changes and the dry run never stops. Filtering is this form.

Research on this pattern

Consequences

  • the actor judges what they wrote by its effect on cases they recognise, rather than by reading it
  • the cases shown decide what the actor learns, so a sample that misses the edge is a false reassurance
  • the run is a rehearsal, and cases that were not in the sample, or arrive later, can still go differently
  • the run produces a before-and-after over cases, which the actor may want to accept or reject one at a time rather than as a whole

Resources & references

Related patterns

Precedes

  • Change review — the run produces a before-and-after over cases, which the actor may want to accept or reject one at a time rather than as a whole

Complements

  • Filtering — a filter over a visible collection is a dry run that never stops; the result updates with each part and the check needs no separate step
  • Plan preview — plan preview shows the system's own plan before a one-off job, dry run shows what the actor's own artifact would do; the direction of authorship is reversed

Related to

  • Action consequences — the severity framework that says when a many-case action warrants a rehearsal before it commits

Preceded by

  • Rule composition — reading the statement does not tell the actor what it matches or does; a selecting rule shows its result on the collection at once, an acting rule shows nothing until it fires