docker-images/linkki-tiedotus/linkki_poster/workflow.py

import re
from dataclasses import dataclass
from pathlib import Path
from typing import Literal

from linkki_poster.formatting import (
    CAPTION_LIMIT,
    ParsedTemplate,
    estimate_caption_length,
    parse_markdown_template,
    render_template,
)
from linkki_poster.models import Config, LANGUAGE_DISPLAY_NAMES, Language, LanguageAssets, PostedMessageRef
from linkki_poster.progress import ProgressDisplay
from linkki_poster.telegram_api import TelegramAPI, TelegramApiError


MESSAGE_LIMIT = 4096
_FAKE_MESSAGE_ID = "9" * 20


@dataclass
class _TextSegmentState:
    label: str
    message_id: int | None
    mode: Literal["message", "caption"]
    parsed_template: ParsedTemplate
    last_rendered: str


@dataclass
class _PostState:
    posted_ref: PostedMessageRef
    text_segments: list[_TextSegmentState]


@dataclass(frozen=True)
class _RuntimeChunkPlan:
    raw_text: str
    parsed_template: ParsedTemplate
    label: str
    has_placeholders: bool


@dataclass
class _LanguageRuntimePlan:
    language_assets: LanguageAssets
    full_template: ParsedTemplate
    full_rendered_text: str
    has_text: bool
    can_use_caption: bool
    chunks: list[_RuntimeChunkPlan]


def post_assets(
    api: TelegramAPI,
    config: Config,
    global_images: list[Path],
    ordered_languages: list[LanguageAssets],
    progress: ProgressDisplay | None = None,
) -> list[PostedMessageRef]:
    reporter = progress or ProgressDisplay()
    leading_global_images, effective_languages = resolve_global_image_strategy(global_images, ordered_languages)
    normalized_username = normalize_channel_username(config.channel_username)
    link_target = resolve_link_target(config.chat_id, normalized_username)
    runtime_plans = build_runtime_plans(effective_languages, link_target)

    posted_message_ids: list[int] = []
    post_states: list[_PostState] = []
    language_links: dict[Language, str] = {}

    try:
        if leading_global_images:
            execute_image_post(
                api=api,
                chat_id=config.chat_id,
                image_paths=leading_global_images,
                label=build_image_label("global", len(leading_global_images)),
                reporter=reporter,
                posted_message_ids=posted_message_ids,
            )

        for runtime_plan in runtime_plans:
            post_state = execute_language_plan(
                api=api,
                config=config,
                runtime_plan=runtime_plan,
                reporter=reporter,
                posted_message_ids=posted_message_ids,
                language_links=language_links,
                link_target=link_target,
            )
            post_states.append(post_state)
            language_links[post_state.posted_ref.language] = post_state.posted_ref.message_url

        execute_final_backfill(api, config.chat_id, post_states, language_links, reporter)
        return [post_state.posted_ref for post_state in post_states]
    except Exception as error:
        rollback_posted_messages(api, config.chat_id, posted_message_ids, reporter)
        raise error


def build_runtime_plans(ordered_languages: list[LanguageAssets], link_target: str) -> list[_LanguageRuntimePlan]:
    runtime_plans: list[_LanguageRuntimePlan] = []
    for language_assets in ordered_languages:
        full_template = parse_markdown_template(language_assets.text_raw)
        full_rendered_text = render_template(full_template, {})
        has_text = full_rendered_text.strip() != ""
        split_chunks = split_markdown_text_for_messages(language_assets.text_raw, link_target) if has_text else []
        chunk_plans = []
        for index, chunk_text in enumerate(split_chunks, start=1):
            parsed_chunk = parse_markdown_template(chunk_text)
            chunk_plans.append(
                _RuntimeChunkPlan(
                    raw_text=chunk_text,
                    parsed_template=parsed_chunk,
                    label=build_text_label(language_assets.language, index, len(split_chunks)),
                    has_placeholders=bool(parsed_chunk.placeholders),
                )
            )
        can_use_caption = (
            bool(language_assets.images)
            and has_text
            and len(chunk_plans) == 1
            and estimate_caption_length(full_template, link_target) <= CAPTION_LIMIT
        )
        runtime_plans.append(
            _LanguageRuntimePlan(
                language_assets=language_assets,
                full_template=full_template,
                full_rendered_text=full_rendered_text,
                has_text=has_text,
                can_use_caption=can_use_caption,
                chunks=chunk_plans,
            )
        )
    return runtime_plans


def execute_language_plan(
    api: TelegramAPI,
    config: Config,
    runtime_plan: _LanguageRuntimePlan,
    reporter: ProgressDisplay,
    posted_message_ids: list[int],
    language_links: dict[Language, str],
    link_target: str,
) -> _PostState:
    language_assets = runtime_plan.language_assets
    language_name = LANGUAGE_DISPLAY_NAMES[language_assets.language]
    text_segments: list[_TextSegmentState] = []
    text_mode: Literal["message", "caption", "none"] = "none"
    text_message_id: int | None = None
    caption_message_id: int | None = None

    if language_assets.images:
        image_label = build_image_label(language_name, len(language_assets.images))

        if runtime_plan.can_use_caption:
            only_chunk = runtime_plan.chunks[0]
            try:
                message_ids = execute_image_post(
                    api=api,
                    chat_id=config.chat_id,
                    image_paths=language_assets.images,
                    label=image_label,
                    reporter=reporter,
                    posted_message_ids=posted_message_ids,
                    caption=render_template(runtime_plan.full_template, language_links),
                )
                caption_message_id = message_ids[0]
                text_mode = "caption"
                text_step_index = reporter.add_step(only_chunk.label, status="done", detail="included as caption")
                text_segments.append(
                    _TextSegmentState(
                        label=only_chunk.label,
                        message_id=caption_message_id,
                        mode="caption",
                        parsed_template=runtime_plan.full_template,
                        last_rendered=render_template(runtime_plan.full_template, language_links),
                    )
                )
            except TelegramApiError as error:
                if not is_caption_too_long_error(error):
                    raise
                reporter.add_step(image_label, status="failed", detail="caption too long, retrying without caption")
                message_ids = execute_image_post(
                    api=api,
                    chat_id=config.chat_id,
                    image_paths=language_assets.images,
                    label=image_label,
                    reporter=reporter,
                    posted_message_ids=posted_message_ids,
                )
                if runtime_plan.has_text:
                    segments, first_message_id = execute_text_chunks(
                        api=api,
                        chat_id=config.chat_id,
                        chunks=runtime_plan.chunks,
                        reporter=reporter,
                        posted_message_ids=posted_message_ids,
                        language_links=language_links,
                    )
                    text_segments.extend(segments)
                    text_message_id = first_message_id
                    text_mode = "message"
                caption_message_id = None
        else:
            message_ids = execute_image_post(
                api=api,
                chat_id=config.chat_id,
                image_paths=language_assets.images,
                label=image_label,
                reporter=reporter,
                posted_message_ids=posted_message_ids,
            )
            if runtime_plan.has_text:
                segments, first_message_id = execute_text_chunks(
                    api=api,
                    chat_id=config.chat_id,
                    chunks=runtime_plan.chunks,
                    reporter=reporter,
                    posted_message_ids=posted_message_ids,
                    language_links=language_links,
                )
                text_segments.extend(segments)
                text_message_id = first_message_id
                text_mode = "message"
    elif runtime_plan.has_text:
        segments, first_message_id = execute_text_chunks(
            api=api,
            chat_id=config.chat_id,
            chunks=runtime_plan.chunks,
            reporter=reporter,
            posted_message_ids=posted_message_ids,
            language_links=language_links,
        )
        text_segments.extend(segments)
        text_message_id = first_message_id
        text_mode = "message"
    else:
        raise ValueError(f"Language '{language_assets.language.value}' has neither text nor images.")

    primary_message_id = caption_message_id or text_message_id
    if primary_message_id is None:
        raise ValueError(f"Language '{language_name}' did not produce a primary message.")

    message_url = build_message_url(link_target, primary_message_id)
    posted_ref = PostedMessageRef(
        language=language_assets.language,
        text_message_id=text_message_id,
        caption_message_id=caption_message_id,
        message_url=message_url,
        had_placeholders=any(bool(segment.parsed_template.placeholders) for segment in text_segments),
        text_mode=text_mode,
    )
    return _PostState(posted_ref=posted_ref, text_segments=text_segments)


def execute_image_post(
    api: TelegramAPI,
    chat_id: str,
    image_paths: list[Path],
    label: str,
    reporter: ProgressDisplay,
    posted_message_ids: list[int],
    caption: str | None = None,
) -> list[int]:
    step_index = reporter.add_step(label)
    reporter.update(step_index, "in_progress")
    response_items = api.send_media_group(
        chat_id,
        image_paths,
        caption=caption,
        parse_mode="HTML" if caption is not None else None,
    )
    message_ids = [int(item["message_id"]) for item in response_items]
    posted_message_ids.extend(message_ids)
    detail = "with caption" if caption is not None else ""
    reporter.update(step_index, "done", detail)
    return message_ids


def execute_text_chunks(
    api: TelegramAPI,
    chat_id: str,
    chunks: list[_RuntimeChunkPlan],
    reporter: ProgressDisplay,
    posted_message_ids: list[int],
    language_links: dict[Language, str],
) -> tuple[list[_TextSegmentState], int]:
    segments: list[_TextSegmentState] = []
    first_message_id: int | None = None

    for chunk in chunks:
        post_step_index = reporter.add_step(chunk.label)
        reporter.update(post_step_index, "in_progress")
        rendered_text = render_template(chunk.parsed_template, language_links)
        detail = ""
        if len(chunks) > 1:
            detail = chunk.label.replace("Post ", "").replace(" text part ", " split part ")
        sent_message = api.send_message(chat_id, rendered_text, parse_mode="HTML")
        message_id = int(sent_message["message_id"])
        posted_message_ids.append(message_id)
        reporter.update(post_step_index, "done", detail)
        if first_message_id is None:
            first_message_id = message_id
        segments.append(
            _TextSegmentState(
                label=chunk.label,
                message_id=message_id,
                mode="message",
                parsed_template=chunk.parsed_template,
                last_rendered=rendered_text,
            )
        )

    if first_message_id is None:
        raise ValueError("Expected at least one text message segment.")
    return segments, first_message_id


def execute_final_backfill(
    api: TelegramAPI,
    chat_id: str,
    post_states: list[_PostState],
    language_links: dict[Language, str],
    reporter: ProgressDisplay,
) -> None:
    for post_state in post_states:
        for text_segment in post_state.text_segments:
            if not text_segment.parsed_template.placeholders:
                continue

            updated_text = render_template(text_segment.parsed_template, language_links)
            if updated_text == text_segment.last_rendered:
                continue

            edit_step_index = reporter.add_step(build_edit_label(text_segment.label))
            reporter.update(edit_step_index, "in_progress")
            if text_segment.message_id is None:
                raise ValueError("Missing message id for edit step.")
            if text_segment.mode == "message":
                api.edit_message_text(chat_id, text_segment.message_id, updated_text, parse_mode="HTML")
            else:
                api.edit_message_caption(chat_id, text_segment.message_id, updated_text, parse_mode="HTML")
            reporter.update(edit_step_index, "done")
            text_segment.last_rendered = updated_text


def rollback_posted_messages(
    api: TelegramAPI,
    chat_id: str,
    posted_message_ids: list[int],
    reporter: ProgressDisplay,
) -> None:
    for message_id in reversed(posted_message_ids):
        step_index = reporter.add_step(f"Delete message {message_id}")
        reporter.update(step_index, "in_progress")
        try:
            api.delete_message(chat_id, message_id)
        except TelegramApiError as error:
            reporter.update(step_index, "failed", error.description)
            continue
        reporter.update(step_index, "done")


def normalize_channel_username(channel_username: str) -> str:
    return channel_username.strip().lstrip("@")


def build_message_url(link_target: str, message_id: int) -> str:
    return f"https://t.me/{link_target}/{message_id}"


def resolve_global_image_strategy(
    global_images: list[Path],
    ordered_languages: list[LanguageAssets],
) -> tuple[list[Path], list[LanguageAssets]]:
    if not global_images:
        return [], ordered_languages

    if not ordered_languages:
        return global_images, ordered_languages

    first_language = ordered_languages[0]
    if first_language.images:
        return global_images, ordered_languages

    with_attached_globals = LanguageAssets(
        language=first_language.language,
        text_file=first_language.text_file,
        text_raw=first_language.text_raw,
        images=global_images,
    )
    return [], [with_attached_globals, *ordered_languages[1:]]


def is_caption_too_long_error(error: TelegramApiError) -> bool:
    return "caption is too long" in error.description.lower()


def split_markdown_text_for_messages(
    markdown_text: str,
    link_target: str,
    message_limit: int = MESSAGE_LIMIT,
) -> list[str]:
    if markdown_text == "":
        return []

    chunks: list[str] = []
    remaining = markdown_text

    while remaining:
        fit_index = find_max_fitting_prefix_index(remaining, link_target, message_limit)
        if fit_index == len(remaining):
            chunks.append(remaining)
            break

        split_index = choose_split_index(remaining, fit_index)
        if split_index <= 0:
            split_index = fit_index

        chunks.append(remaining[:split_index])
        remaining = remaining[split_index:]

    return [chunk for chunk in chunks if chunk != ""]


def find_max_fitting_prefix_index(markdown_text: str, link_target: str, message_limit: int) -> int:
    cached_lengths: dict[int, int] = {}

    def measured_length(end_index: int) -> int:
        if end_index not in cached_lengths:
            cached_lengths[end_index] = estimate_rendered_message_length(markdown_text[:end_index], link_target)
        return cached_lengths[end_index]

    if measured_length(len(markdown_text)) <= message_limit:
        return len(markdown_text)

    low = 1
    high = len(markdown_text)
    best = 0

    while low <= high:
        middle = (low + high) // 2
        if measured_length(middle) <= message_limit:
            best = middle
            low = middle + 1
        else:
            high = middle - 1

    if best == 0:
        raise ValueError("Unable to split message content into Telegram-compatible chunks.")
    return best


def choose_split_index(markdown_text: str, max_index: int) -> int:
    double_newline_match = find_last_break_end(markdown_text, re.compile(r"\n{2,}"), max_index)
    if double_newline_match is not None:
        return double_newline_match

    single_newline_index = markdown_text.rfind("\n", 0, max_index)
    if single_newline_index != -1:
        return single_newline_index + 1

    whitespace_index = find_last_whitespace_break(markdown_text, max_index)
    if whitespace_index is not None:
        return whitespace_index

    return max_index


def find_last_break_end(markdown_text: str, pattern: re.Pattern[str], max_index: int) -> int | None:
    last_end: int | None = None
    for match in pattern.finditer(markdown_text):
        if match.end() > max_index:
            break
        last_end = match.end()
    return last_end


def find_last_whitespace_break(markdown_text: str, max_index: int) -> int | None:
    for index in range(max_index - 1, -1, -1):
        if markdown_text[index] in {" ", "\t"}:
            return index + 1
    return None


def estimate_rendered_message_length(markdown_text: str, link_target: str) -> int:
    parsed_template = parse_markdown_template(markdown_text)
    fake_url = f"https://t.me/{link_target}/{_FAKE_MESSAGE_ID}"
    fake_links = {
        Language.FI: fake_url,
        Language.SV: fake_url,
        Language.EN: fake_url,
    }
    rendered_text = render_template(parsed_template, fake_links)
    return len(rendered_text)


def resolve_link_target(chat_id: str, normalized_username: str = "") -> str:
    if normalized_username != "":
        return normalized_username

    internal_chat_id = extract_internal_chat_id(chat_id)
    if internal_chat_id is None:
        raise ValueError(
            "Cannot generate message links without channel_username or a channel-style chat_id (-100...)."
        )
    return f"c/{internal_chat_id}"


def extract_internal_chat_id(chat_id: str) -> str | None:
    stripped = chat_id.strip()
    if not stripped.startswith("-100"):
        return None
    suffix = stripped[4:]
    if suffix == "" or not suffix.isdigit():
        return None
    return suffix


def build_image_label(target: str, image_count: int) -> str:
    noun = "image" if image_count == 1 else "images"
    return f"Post {image_count} {target} {noun}"


def build_text_label(language: Language, chunk_index: int, chunk_count: int) -> str:
    language_name = LANGUAGE_DISPLAY_NAMES[language]
    if chunk_count == 1:
        return f"Post {language_name} text"
    return f"Post {language_name} text part {chunk_index}/{chunk_count}"


def build_edit_label(text_label: str) -> str:
    if text_label.startswith("Post "):
        return "Edit " + text_label[5:]
    return "Edit " + text_label