Module edit 

Shared edit engine: byte-offset conversion, content replacement, syntax validation, and auto-backup orchestration.

Used by write, edit_symbol, edit_match, and batch commands.

Structs

ReformattedExcerpt

Post-format excerpt of the region(s) the formatter reflowed, so the agent can re-anchor its next edit on the real on-disk text instead of the text it submitted. None on WriteResult when the formatter did not change the applied edit (self-suppressing).

WriteResult

Result of the write → format → validate pipeline.

Functions

attach_preview_diff

Attach the standard preview diff fields to a command response payload.

auto_backup

Snapshot the file into the backup store before mutation, scoped to a session.

build_unified_diff

Build the unified-diff string used by internal permission previews.

compute_diff_counts

Compute compact diff counts (additions/deletions) without echoing any file content. This is the agent-facing default — the payload is constant-size regardless of how large the edited file is.

compute_diff_for_response

Pick the right diff shape for a response based on request flags.

compute_diff_info

Compute diff info between before/after content for UI metadata. Returns a JSON value with before, after, additions, deletions. For files >512KB, omits full content and returns only counts.

compute_reformatted_excerpt

Compute a bounded post-format excerpt when pre_format (agent-applied edit) differs from post_format (on-disk after formatting).

format_validation_errors

Render structured validation errors as a compact line N: message list for an error message. Used by the refactor handlers when a write was rolled back.

line_col_to_byte

Convert 0-indexed line/col to a byte offset within source.

replace_byte_range

Replace bytes in [start..end) with replacement.

validate_syntax

Validate syntax of a file using a fresh FileParser (D023).

validate_syntax_str

Validate syntax of an in-memory string without touching disk.

wants_diff

Check if the caller requested diff info in the response.

wants_diff_content

Check if the caller requested the full before/after file contents in the diff. This is for UI rendering only (e.g. the OpenCode/Pi plugins building a diff view in tool metadata) and is deliberately NOT the default: full content makes the response scale with file size, not edit size, which floods agent context on large files. Agent-facing/raw consumers should pass include_diff (counts only) instead.

wants_preview

Check whether the caller requested an internal, side-effect-free preview.

write_format_validate

Write content to disk, auto-format, then validate syntax.