Struct ReplayReport 

pub struct ReplayReport {
    pub header: SessionHeader,
    pub total_entries: usize,
    pub applied: usize,
    pub skipped: Vec<(usize, String)>,
    pub stopped_at: Option<usize>,
    pub not_applied_after_stop: usize,
    pub second_session_at: Option<usize>,
    pub deterministic: bool,
    pub live_fingerprint: Option<String>,
    pub matches_live: Option<bool>,
    pub state: State,
}

Everything --replay learned from one recording.

Fields

header: SessionHeader

total_entries: usize

Entry lines seen for this session (excludes the header line).

applied: usize

Entries successfully reconstructed and folded through update.

skipped: Vec<(usize, String)>

(file line number, reason) for every line that could not be reconstructed — malformed JSON, or a Msg variant this build doesn’t know (log written by a newer mermaid).

stopped_at: Option<usize>

File line number where the fold stopped because the recorded session quit (state.should_exit), mirroring the live loop’s exit.

not_applied_after_stop: usize

Entries after the stop line that were parsed but not applied.

second_session_at: Option<usize>

Line number of a second SessionHeader if the file holds multiple appended sessions. Only the first session is replayed.

deterministic: bool

True when folding the log twice produced byte-identical State debug representations — the reducer-purity invariant.

live_fingerprint: Option<String>

The live session’s final fingerprint, when the recording was sealed on clean exit. None for crashed sessions and pre-trailer logs.

matches_live: Option<bool>

Whether the folded session reproduces the live one: Some(true) = exact match, Some(false) = diverged (expected when redaction fired mid-session or the reducer changed since recording), None = no trailer to compare against.

state: State

Final state after the fold.