Fixing what it finds
Every finding names the rules behind it, and you can ask what would happen if one were switched off — without switching anything off.
A finding that says "this is impossible" and stops there is only half useful. The question you actually have is which rule do I change.
Every finding names its rules
Open a finding that rules produced, and it lists those constraints by their own identifying names rather than as internal references. Select a rule name and that rule opens in a panel on top of the report. The panel gives the rule's weight, whether it is switched on, and every teacher, student group, room, day, hour and activity it names.
You stay on the report. Reading a rule no longer costs you your place in the findings.
On the data screen the panel also offers Open in the table, which takes you to that rule in the constraints list. The public checker has no constraints table, so the panel does not offer that there.
For an unplaceable lesson the report also shows how it got there — the number of starting times it began with, and how many were left after each rule, in order. That trace is usually where the answer is: one step in it is almost always a rule that is stricter than you meant.
The culprits are minimised
When a lesson has been left with no legal starting time, the check does not just list every rule that touched it. It tries to work out the smallest set of rules that still causes the problem.
That matters when a lesson is covered by a dozen constraints. Being told "these two rules, together, are what makes this impossible" is actionable; being handed a list of twelve is not.
The report tells you when it succeeded. A narrowed list carries the words only the rules that actually cause it. Without those words the list is every rule that touched the lesson, and the work of narrowing it is yours. The check has a budget, so a very large file can run out before it finishes narrowing.
"What if a rule were off?"
⚠ This is on the data screen only. The public checker at fetviewer.com/check does not offer it.
A blocking finding has a What if a rule were off? button. It re-runs the whole check once for each rule the finding names, with that rule left out. For each rule, it shows how many blocking problems the file would still have. It does not say which findings those are.
Nothing is saved and nothing is switched off. It is a question, not an edit. Your rules are exactly as they were when you close it.
This is the fastest way through a tangle: rather than disabling a rule, running FET, and undoing it, you ask each candidate in turn and change the one that actually accounts for the problem.
Working order
- Clear Proven impossible findings first. Until they are gone, the solve cannot succeed, so anything else is polish.
- Use the trace to find the strictest step rather than the most recent rule.
- Ask "What if a rule were off?" before editing anything.
- Re-run the check. The report is regenerated from your current rules, so it answers immediately.
- Then look at Mistakes. Dangling references and duplicates are cheap to fix, and they stop next term's confusion.