core/ui/transcript_view.py

"""Speaker chip palette, styler, and grouped-row time mapping for transcript DataFrame."""

import re

import pandas as pd


SPEAKER_PALETTE = [
    "#FFD23F",  # yellow
    "#FF6B6B",  # coral
    "#5CE1E6",  # cyan
    "#A4F47A",  # lime
    "#FF6FC0",  # magenta
    "#FF9F1C",  # orange
]


def _extract_index(label: str) -> int:
    """Pull a stable integer index out of a speaker label.

    SPEAKER_00 → 0, SPK-1 → 1, "Alice" → hash-fallback so renames keep deterministic colors
    only when caller passes original label; once renamed without a digit, fall back to hash.
    """
    if not label:
        return 0
    m = re.search(r"(\d+)", str(label))
    if m:
        return int(m.group(1))
    return abs(hash(str(label))) % len(SPEAKER_PALETTE)


def color_for_speaker(label: str) -> str:
    return SPEAKER_PALETTE[_extract_index(label) % len(SPEAKER_PALETTE)]


def render_chip(label: str) -> str:
    color = color_for_speaker(label)
    return f'<span class="brut-chip" style="background:{color}">▣ {label}</span>'


def style_transcript(df: pd.DataFrame):
    """Return a Styler that paints the Speaker column cell background per palette."""
    if df is None or df.empty or "Speaker" not in df.columns:
        return df

    def _color(value):
        c = color_for_speaker(value)
        return f"background-color: {c}; font-family: 'Archivo Black', sans-serif;"

    try:
        return df.style.applymap(_color, subset=["Speaker"])
    except AttributeError:
        # pandas >= 2.1 deprecates applymap → use map
        return df.style.map(_color, subset=["Speaker"])


def grouped_row_start_seconds(merged: list, row_idx: int) -> float:
    """Walk merged segments, group consecutive same-speaker turns, return start seconds of group at row_idx."""
    if not merged or row_idx is None or row_idx < 0:
        return 0.0

    group_idx = -1
    prev_speaker = None
    for seg in merged:
        spk = seg.get("speaker")
        if spk != prev_speaker:
            group_idx += 1
            if group_idx == row_idx:
                return float(seg.get("start", 0.0))
            prev_speaker = spk
    return 0.0


def build_label_map(edited_df, merged: list) -> dict:
    """From edited DataFrame Speaker column + original merged segments, build {orig_label → new_label}.

    Walks grouped segments in order, mapping each group's original speaker label to the
    corresponding row's edited Speaker value.
    """
    if edited_df is None or merged is None:
        return {}

    if hasattr(edited_df, "values"):
        rows = edited_df.values.tolist() if isinstance(edited_df, pd.DataFrame) else edited_df
        if isinstance(edited_df, pd.DataFrame) and "Speaker" in edited_df.columns:
            new_labels = edited_df["Speaker"].tolist()
        else:
            new_labels = [r[2] for r in rows if len(r) > 2]
    elif isinstance(edited_df, list):
        new_labels = [r[2] for r in edited_df if len(r) > 2]
    else:
        return {}

    label_map: dict = {}
    group_idx = -1
    prev_speaker = None
    for seg in merged:
        spk = seg.get("speaker")
        if spk != prev_speaker:
            group_idx += 1
            if group_idx < len(new_labels):
                new_label = str(new_labels[group_idx]).strip()
                if new_label and new_label != spk:
                    label_map[spk] = new_label
            prev_speaker = spk
    return label_map