An import that reports fifteen failures and can name ten

[dmitrat]

Phase 7 of the fix plan: import and export. Studio 1000 green, 989 before.

Finding 27 was half built, which is why it was reported as absent

The CSV path has kept every rejection since WS-36, and the window has a button that writes them
beside the source with the line, the reason and the text. The JSON path filled the display list and
recorded nothing — so a JSON import really did report 15 failed, name ten, and have nothing to
save. Both paths record now, and the rule counts recording sites against display sites rather than
trusting either.

finding	fix
25	0 columns for every file, because the count asked ColumnMappings — built on step 3. It reads the parse now, and the rows themselves are drawn: a wrong delimiter is obvious in the rows and invisible in a count
26	the preview banner and the error banner were both Grid.Row="5", so the result landed across the line saying what the file held. Two borders in one cell are drawn over each other at full opacity, in every state — the same shape as the conflict message in Phase 4
28	the collision policy moved to the step it is about; Columns held two unrelated decisions while Destination held one control
32	the export's first frame said 0 / 0 rows at 0.0 %, because the panel opens before the rows are fetched. The numbers wait for the total
4	the output path can be typed as well as browsed for
5	Export table and Export query results, in the title bar and in the Tools menu — only one of them carries the three scopes with their counts

The census of hard-coded values went red again at 69 against 68 (the preview's MaxHeight); the
ledger records it with its reason — a box round a block of content, so the wizard does not grow with
the file.

🤖 Generated with Claude Code