demo/src/redaction.py

from __future__ import annotations

import hashlib
import html
import re

from src.domain import PolicyAction, RULESET_ID
from src.normalize import fold_digits
from src.schemas import Span


def fingerprint(value: str) -> str:
    digest = hashlib.sha256(f"{RULESET_ID}:{value}".encode("utf-8")).hexdigest()
    return digest[:10]


def mask_token(span: Span) -> str:
    return f"[{span.entity_type.value}:{fingerprint(span.surface)}]"


def apply_text(text: str, spans: list[Span], action: PolicyAction) -> str:
    if not text:
        return ""
    ordered = sorted(spans, key=lambda item: item.start)
    chunks: list[str] = []
    cursor = 0
    for item in ordered:
        if item.start < cursor or item.end > len(text) or item.start >= item.end:
            continue
        chunks.append(text[cursor : item.start])
        chunks.append("" if action is PolicyAction.REMOVE else mask_token(item))
        cursor = item.end
    chunks.append(text[cursor:])
    return "".join(chunks)


def leftover_values(export_text: str, values: list[str]) -> list[str]:
    haystack = fold_digits(export_text or "")
    leaked: list[str] = []
    for value in values:
        needle = fold_digits(value or "")
        if needle and len(needle) >= 4 and needle in haystack:
            leaked.append(value)
    return leaked


def highlight_html(text: str, spans: list[Span], rtl: bool = False) -> str:
    direction = "rtl-mode" if rtl else ""
    if not text:
        return f"<pre class='doc-view {direction}'></pre>"
    ordered = sorted(spans, key=lambda item: (item.start, item.end))
    chunks: list[str] = []
    cursor = 0
    for item in ordered:
        if item.start < cursor or item.end > len(text):
            continue
        chunks.append(html.escape(text[cursor : item.start]))
        title = html.escape(f"{item.entity_type.value} {item.confidence:.2f}")
        chunks.append(f"<mark class='pii-span' title='{title}'>{html.escape(text[item.start:item.end])}</mark>")
        cursor = item.end
    chunks.append(html.escape(text[cursor:]))
    return f"<pre class='doc-view {direction}'>{''.join(chunks)}</pre>"


def redact_csv_cells(text: str, spans: list[Span], action: PolicyAction) -> str:
    return apply_text(text, spans, action)


TOKEN_RE = re.compile(r"\[[A-Z_]+:[0-9a-f]{10}\]")


def highlight_redacted_html(text: str, rtl: bool = False) -> str:
    direction = "rtl-mode" if rtl else ""
    escaped = html.escape(text or "")
    marked = TOKEN_RE.sub(lambda match: f"<mark class='redact-span'>{match.group(0)}</mark>", escaped)
    return f"<pre class='doc-view {direction}'>{marked}</pre>"