app.py

# app.py
import os
import io
import sys
import zipfile
import tempfile
import subprocess
from pathlib import Path

import gradio as gr
import pandas as pd

import barzooka as bz


def _check_pdftocairo() -> str:
    try:
        out = subprocess.run(["pdftocairo", "-v"], stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
        if out.returncode == 0 or "pdftocairo" in out.stdout:
            return out.stdout.strip().splitlines()[-1]
        return f"pdftocairo not found (returncode {out.returncode}). Output:\n{out.stdout}"
    except FileNotFoundError:
        return "pdftocairo not found (FileNotFoundError). Install poppler-utils."


def _maybe_download_model(target_path: Path) -> bool:
    """If BARZOOKA_MODEL_URL is set, try to download the model to target_path."""
    url = os.getenv("BARZOOKA_MODEL_URL", "").strip()
    if not url:
        return False
    try:
        import urllib.request
        print(f"Downloading Barzooka model from {url} ...")
        with urllib.request.urlopen(url) as resp:
            data = resp.read()
        target_path.write_bytes(data)
        print(f"Saved model to {target_path}")
        return True
    except Exception as e:
        print(f"Model download failed: {e}")
        return False


def _construct_barzooka() -> bz.Barzooka:
    # 1) Try package-embedded resource
    try:
        return bz.Barzooka(cpu=True)
    except Exception as e_default:
        # 2) Fallback: local pkl next to app.py
        local_pkl = Path(__file__).with_name("barzooka.pkl")
        if not local_pkl.exists():
            # 2a) Try to download if env var provided
            _maybe_download_model(local_pkl)
        if local_pkl.exists():
            try:
                return bz.Barzooka(model_file=str(local_pkl), cpu=True)
            except Exception as e_local:
                raise RuntimeError(
                    "Failed to load Barzooka model from both package resources and local 'barzooka.pkl'.\n"
                    f"Package error: {e_default}\nLocal file error: {e_local}"
                )
        raise RuntimeError(
            "Barzooka model file not found. Upload 'barzooka.pkl' next to app.py, or set BARZOOKA_MODEL_URL to a downloadable .pkl."
        )


BARZOOKA = _construct_barzooka()


def _predict_single_pdf(pdf_path: str, pagewise: bool = False) -> pd.DataFrame:
    res = BARZOOKA.predict_from_file(pdf_path, tmp_folder="./tmp/", pagewise=pagewise)
    if pagewise:
        images, page_preds = res
        rows = [{"page": i + 1, "classes": preds} for i, preds in enumerate(page_preds)]
        return pd.DataFrame(rows)
    else:
        return pd.DataFrame([res])


def screen_pdf(pdf_file, pagewise: bool):
    if pdf_file is None:
        return pd.DataFrame([{"error": "Please upload a PDF."}])
    try:
        return _predict_single_pdf(pdf_file, pagewise)
    except Exception as e:
        return pd.DataFrame([{"error": str(e)}])


def screen_images(img_files):
    if not img_files:
        return pd.DataFrame([{"error": "Please upload at least one image."}])
    try:
        classes_per_img = BARZOOKA.predict_from_img(img_files)
        rows = [{"image": os.path.basename(p), "classes": c} for p, c in zip(img_files, classes_per_img)]
        return pd.DataFrame(rows)
    except Exception as e:
        return pd.DataFrame([{"error": str(e)}])


def screen_zip_of_pdfs(zip_file, pagewise: bool):
    if zip_file is None:
        return pd.DataFrame([{"error": "Please upload a .zip containing PDFs."}]), None

    out_rows = []

    with tempfile.TemporaryDirectory() as tdir:
        with zipfile.ZipFile(zip_file, "r") as zf:
            zf.extractall(tdir)

        pdf_paths = [str(p) for p in Path(tdir).rglob("*.pdf")]
        if not pdf_paths:
            return pd.DataFrame([{"error": "No PDFs found in the ZIP."}]), None

        for pdf in sorted(pdf_paths):
            try:
                df = _predict_single_pdf(pdf, pagewise=pagewise)
                if pagewise:
                    paper_id = Path(pdf).stem
                    df.insert(0, "paper_id", paper_id)
                    out_rows.append(df)
                else:
                    out_rows.append(df)
            except Exception as e:
                out_rows.append(pd.DataFrame([{"paper_id": Path(pdf).stem, "error": str(e)}]))

        result_df = pd.concat(out_rows, ignore_index=True)

        csv_path = Path(tdir) / "barzooka_results.csv"
        result_df.to_csv(csv_path, index=False)

        out_csv = Path.cwd() / "barzooka_results.csv"
        try:
            out_csv.write_bytes(csv_path.read_bytes())
        except Exception:
            out_csv = None

        return result_df, (str(out_csv) if out_csv and out_csv.exists() else None)


def diagnostics():
    pdfto = _check_pdftocairo()
    classes = getattr(BARZOOKA, "class_names", [])
    summary = {
        "pdftocairo": pdfto,
        "classes": classes,
        "model_loaded": True,
    }
    return pd.DataFrame([summary])


with gr.Blocks(title="Barzooka: Graph-Type Screening") as demo:
    gr.Markdown(
        """
        # Barzooka
        Screen PDFs or images to detect graph types (bar, bardot, box, dot, violin, hist, pie, flow, text, other).

        **Notes:** PDF screening requires the system utility `pdftocairo` (from Poppler). This Space installs
        `poppler-utils` so conversion should work out-of-the-box.
        """
    )

    with gr.Row():
        diag_btn = gr.Button("Environment diagnostics")
        diag_out = gr.Dataframe(label="Diagnostics", wrap=True)
        diag_btn.click(diagnostics, outputs=diag_out)

    with gr.Tab("Single PDF"):
        pdf_input = gr.File(label="Upload a PDF", file_types=[".pdf"])
        pagewise_chk = gr.Checkbox(label="Return page-wise predictions", value=False)
        pdf_btn = gr.Button("Screen PDF")
        pdf_out = gr.Dataframe(label="Results", wrap=True)
        pdf_btn.click(fn=screen_pdf, inputs=[pdf_input, pagewise_chk], outputs=pdf_out)

    with gr.Tab("ZIP of PDFs (batch)"):
        zip_input = gr.File(label="Upload a ZIP containing PDFs", file_types=[".zip"])
        pw_chk = gr.Checkbox(label="Return page-wise predictions", value=False)
        zip_btn = gr.Button("Screen ZIP")
        zip_out = gr.Dataframe(label="Results (preview)", wrap=True)
        zip_csv = gr.File(label="Download all results (CSV)")
        zip_btn.click(fn=screen_zip_of_pdfs, inputs=[zip_input, pw_chk], outputs=[zip_out, zip_csv])

    with gr.Tab("Images"):
        imgs_input = gr.Files(label="Upload images", file_types=[".jpg", ".jpeg", ".png"])
        imgs_btn = gr.Button("Screen Images")
        imgs_out = gr.Dataframe(label="Results", wrap=True)
        imgs_btn.click(fn=screen_images, inputs=imgs_input, outputs=imgs_out)

demo.queue(api_open=True)


#In Spaces, you don’t need to call launch(); for local testing:
if __name__ == "__main__":
    demo.queue(api_open=True).launch()