app.py

import re
import json
import requests
import matplotlib.pyplot as plt
import gradio as gr
from requests.exceptions import HTTPError


def parse_roboflow_url(url):
    """
    Extract workspace/project and version from a Roboflow Universe URL.
    Returns (workspace, project, version)
    """
    pattern = r"roboflow\.com/([^/]+)/([^/]+)/dataset/(\d+)"
    match = re.search(pattern, url)
    if not match:
        raise ValueError(f"Invalid Roboflow dataset URL: {url}")
    return match.groups()


def fetch_metadata(api_key, workspace, project, version):
    """
    Fetch metadata from Roboflow. Raises ValueError on HTTP errors.
    """
    endpoint = f"https://api.roboflow.com/{workspace}/{project}/{version}"
    resp = requests.get(endpoint, params={"api_key": api_key})
    try:
        resp.raise_for_status()
    except HTTPError:
        if resp.status_code == 401:
            raise ValueError("Unauthorized: check your API key.")
        else:
            raise ValueError(f"Error {resp.status_code} for {workspace}/{project}/{version}")
    data = resp.json()
    total = data.get("version", {}).get("images") or data.get("project", {}).get("images", 0)
    classes = data.get("project", {}).get("classes", {})
    return total, classes


def aggregate_datasets(api_key, entries):
    """
    Given list of (url, file, line), returns:
      - total_images
      - dict[class_name_lowercase] = aggregated count
      - dict[class_name_lowercase] = set(source URLs)
    """
    total_images = 0
    class_counts = {}
    class_sources = {}
    for url, fname, lineno in entries:
        try:
            ws, proj, ver = parse_roboflow_url(url)
        except ValueError:
            raise ValueError(f"Invalid URL '{url}' in file '{fname}', line {lineno}")
        imgs, cls_map = fetch_metadata(api_key, ws, proj, ver)
        total_images += imgs
        for cls, cnt in cls_map.items():
            norm = cls.strip().lower()
            class_counts[norm] = class_counts.get(norm, 0) + cnt
            class_sources.setdefault(norm, set()).add(url)
    return total_images, class_counts, class_sources


def make_bar_chart(counts):
    """
    Build a bar chart from a {label: value} dict.
    """
    fig, ax = plt.subplots()
    labels = list(counts.keys())
    values = list(counts.values())
    ax.bar(range(len(labels)), values)
    ax.set_xticks(range(len(labels)))
    ax.set_xticklabels(labels, rotation=45, ha="right")
    ax.set_ylabel("Image Count")
    ax.set_title("Class Distribution")
    fig.tight_layout()
    return fig


def load_datasets(api_key, file_objs):
    """
    1) Ensure API key present
    2) Read & dedupe URLs from each uploaded .txt
    3) Fetch & aggregate metadata
    Returns: total, table_data, figure, json_counts, markdown_sources
    """
    if not api_key or not api_key.strip():
        raise ValueError("Please enter your Roboflow API Key before loading datasets.")

    entries = []
    seen = set()
    for fobj in file_objs:
        fname = getattr(fobj, "name", None) or fobj.get("name", "unknown")
        # read raw bytes or dict-data or file path
        try:
            raw = fobj.read()
        except:
            raw = fobj.get("data") if isinstance(fobj, dict) else None
        if raw is None and isinstance(fobj, str):
            with open(fobj, "rb") as fh:
                raw = fh.read()
        text = raw.decode("utf-8") if isinstance(raw, (bytes, bytearray)) else raw

        for i, line in enumerate(text.splitlines(), start=1):
            url = line.strip()
            if url and url not in seen:
                seen.add(url)
                entries.append((url, fname, i))

    total, counts, sources = aggregate_datasets(api_key, entries)

    # build dataframe rows
    table_data = [[cls, counts[cls]] for cls in counts]

    # build clickable markdown per-class
    md_lines = []
    for cls in counts:
        links = ", ".join(f"[{url.split('/')[-1]}]({url})" for url in sources[cls])
        md_lines.append(f"- **{cls}** ({counts[cls]} images): {links}")
    md_sources = "\n".join(md_lines)

    fig = make_bar_chart(counts)
    return str(total), table_data, fig, json.dumps(counts, indent=2), md_sources


def update_classes(df_data):
    """
    Convert df_data into a list-of-lists (if needed),
    merge duplicate/lowercased class names, and recalc all outputs.
    Returns: total, updated_table, figure, json_counts, markdown_summary
    """
    # convert Pandas DataFrame or NumPy array into list-of-lists
    if not isinstance(df_data, list):
        if hasattr(df_data, "to_numpy"):
            df_data = df_data.to_numpy().tolist()
        elif hasattr(df_data, "tolist"):
            df_data = df_data.tolist()

    combined = {}
    for row in df_data:
        if len(row) < 2:
            continue
        name, cnt = row[0], row[1]
        if not name:
            continue
        key = str(name).strip().lower()
        try:
            val = int(cnt)
        except:
            val = 0
        combined[key] = combined.get(key, 0) + val

    total = sum(combined.values())
    updated_table = [[k, combined[k]] for k in combined]
    fig = make_bar_chart(combined)
    md_summary = "\n".join(f"- **{k}** ({combined[k]} images)" for k in combined)

    return str(total), updated_table, fig, json.dumps(combined, indent=2), md_summary


def build_ui():
    with gr.Blocks() as demo:
        gr.Markdown("## Roboflow Dataset Inspector")

        with gr.Row():
            api_input = gr.Textbox(label="API Key", type="password")
            files = gr.Files(label="Upload .txt files", file_types=[".txt"])

        load_btn = gr.Button("Load Datasets")
        total_out = gr.Textbox(label="Total Images", interactive=False)
        df = gr.Dataframe(
            headers=["Class Name", "Count"],
            row_count=(1, None),
            col_count=2,
            interactive=True
        )
        plot = gr.Plot()
        json_out = gr.Textbox(label="Counts (JSON)", interactive=False)
        md_out = gr.Markdown(label="Class Sources")

        update_btn = gr.Button("Apply Class Edits")

        load_btn.click(
            fn=load_datasets,
            inputs=[api_input, files],
            outputs=[total_out, df, plot, json_out, md_out]
        )
        update_btn.click(
            fn=update_classes,
            inputs=[df],
            outputs=[total_out, df, plot, json_out, md_out]
        )

    return demo


if __name__ == "__main__":
    build_ui().launch()