Replace Text Within One File After Classifying Ten Matches

A file replacement workflow that checks targets and exceptions first

Safe replacement within one file begins with a semantic split, not Replace All. The controlled chapter contains ten instances of “totemo”: eight narrative uses should become “hijō ni,” while two dialogue uses remain as voice. The operation runs only on a duplicate of chapter 5. Its final evidence is two old-form exceptions and eight new forms, not a zero count across the workspace.

Classify ten matches before opening replacement

Duplicate the 3,200-character chapter and search for the old phrase. During the first pass, keep replacement closed and label eight narrative matches N1 through N8 and two dialogue matches D1 and D2.

Dialogue is not automatically protected. These two are exceptions because the character’s emphatic voice is an explicit editorial decision.

Read one sentence around each adverb

Confirm that replacing the narrative adverb preserves the sentence meaning. Preserve the repeated dialogue form where repetition carries voice. Surface equality does not make the ten contexts editorially equal.

If a proposed new form is wrong for one narrative sentence, remove that item from the replacement set rather than forcing the expected count.

Turn navigation order into a decision list

Record number, short context, narrative or dialogue, replace or preserve, and reason. Add a short opening fragment beside each line number so later line movement does not destroy traceability. Confirm N1 through N8 are targets and D1 through D2 are exceptions.

Replace eight matches one at a time

Enter the old and new terms, check the current occurrence, and execute only when it is N1 through N8. Skip D1 and D2. Ten navigation steps should produce eight changes.

Replace All cannot express the voice exception, and this article does not state that a regular expression or application understands narrative meaning.

Keep scope on the chapter-5 duplicate

Name the file chapter05_replace-test and keep other chapter tabs outside the operation. Occurrences in other chapters are not added to the ten. Multi-file scope and per-file totals belong to the bounded multi-file replacement workflow.

Reverse-search for two old-form exceptions

After eight changes, search the old term. The expected result is two, not zero. Navigate to D1 and D2 and compare each with its recorded reason.

Three old forms mean one target was skipped. One old form means an exception was changed. The count determines the return point.

Search for eight new forms

The controlled specimen begins with zero new forms, so the final expectation is eight. Nine suggests a pre-existing or accidental form; seven suggests a missed or unsaved change. Without the starting count, eight after replacement would not be sufficient evidence.

Check a diff for exactly eight word changes

Compare the duplicate with the original chapter. Only N1 through N8 should change. Punctuation, line breaks, quotation marks, and D1 through D2 remain identical. Matching counts do not excuse surrounding damage.

Separate confirmed replacement results from the untested interface

Rune Studio documentation describes Cmd+F local search, using selected text as the query, moving among matches, and working with search and replacement fields.

On August 14, 2026, a hands-on check with a disposable copy searched one file, recorded the match count, ran a preview of planned changes, replaced the planned matches, and reverse-searched the old and new terms. Single-file search, preview of planned changes, replacement, and reverse search have been confirmed in a hands-on check.

The Cmd+F interface, selected-text handoff, previous/next controls, and the article-specific ten-to-eight-plus-two specimen were not observed. The 8/2 count is an acceptance-test expectation, not an automatic classification result.

Record zero starting new forms

Immediately after counting ten old forms, search the new form and record zero in the same chapter duplicate. An existing new form would change the final expectation, so old and new starting values are separate evidence.

Identify D1 and D2 with quoted context

Record the two dialogue exceptions with quotation marks and short sentence fragments. A line number alone is unstable, and quotation marks alone do not prove dialogue; speaker and editorial role still decide the D label.

Undo an accidental D1 replacement directly

If D1 is changed, undo that occurrence and verify restoration before continuing. Do not leave the mistake and skip an N item merely to make the total eight. Preserve the correction event in the record.

Reopen the saved duplicate and verify 8/2

Save, close, and reopen the duplicate, then search for eight new and two old forms again. This separates editor actions from persisted file state. Only the reopened 8/2 result is handed toward the master.

Exclude substring matches from the ten semantic uses

Check whether every result is an independent adverb rather than a fragment inside a longer token or note marker. The controlled ten are independent uses; any substring result returns to query or classification.

Read the entire chapter after N1 through N8

A numerical 8/2 result can still make the narrative unnaturally formal. Read from chapter start to end and reconsider any N item that disrupts voice. Update the expected count if a justified target is withdrawn.

Hand master integration a position-aware list

For N1 through N8, provide context fragment, old term, new term, and chapter position. The master editor chooses controlled replacement or diff integration; the duplicate is not renamed over the master without a separate identity check.

Close replacement before the final search-only pass

Close the replacement field, navigate only the two old exceptions, then navigate the eight new forms. This prevents a final confirmation click from becoming another edit. Compare every position with D1–D2 and N1–N8 before closing the duplicate.

Close with eight changes and two preserved uses

Final evidence is two old forms at D1 and D2, eight new forms at N1 through N8, an eight-word diff, and one target file. Do not pursue zero old forms.

Duplicate chapter 5 and review the local search and replacement boundary on the Rune Studio product page. Rebuild scope in the bounded multi-file replacement workflow before any cross-file operation.