# 常用
pip install pypdf pdfminer.six pdfplumber pymupdf

# 表格
pip install "camelot-py[base]"     # 1.0+ 默認用 PDFium，無需 Ghostscript（Linux 下更易裝）:contentReference[oaicite:8]{index=8}
pip install tabula-py               # 需 Java 8+ 運行環(huán)境 :contentReference[oaicite:9]{index=9}

# OCR 路線1：一站式
# macOS 可: brew install ocrmypdf ；Linux/Win 請看文檔 :contentReference[oaicite:10]{index=10}

# OCR 路線2：Python 組合
pip install pdf2image pytesseract   # 還需安裝 Poppler 與 Tesseract 可執(zhí)行文件 :contentReference[oaicite:11]{index=11}

# 深入與底層
pip install pypdfium2 pikepdf       # PDFium 綁定 & 元數(shù)據(jù)/結構 :contentReference[oaicite:12]{index=12}

from pypdf import PdfReader

reader = PdfReader("input.pdf")
text = "\n".join((page.extract_text() or "") for page in reader.pages)
print(text)

from pdfminer.high_level import extract_pages
from pdfminer.layout import LTTextContainer, LTChar, LAParams

for page_layout in extract_pages("input.pdf", laparams=LAParams()):
    for element in page_layout:
        if isinstance(element, LTTextContainer):
            print(element.get_text().strip())

import pdfplumber

with pdfplumber.open("input.pdf") as pdf:
    for i, page in enumerate(pdf.pages, 1):
        # 文本（可調 x/y 容差，提取更平滑）
        t = page.extract_text(x_tolerance=1, y_tolerance=3) or ""
        print(f"--- Page {i} ---\n{t}\n")

        # 表格（簡單嘗試）
        for table in page.extract_tables():
            for row in table:
                print(row)

import fitz  # PyMuPDF

doc = fitz.open("input.pdf")
for page in doc:
    # “text”=純文本；“blocks”=文本塊；“dict/json/html”=結構化/富文本輸出
    print(page.get_text("blocks"))
    links = page.get_links()          # 鏈接
    annots = [a.info for a in page.annots() or []]  # 注釋

import fitz
doc = fitz.open("input.pdf")
for page_index, page in enumerate(doc):
    for img in page.get_images(full=True):
        xref = img[0]
        pix = fitz.Pixmap(doc, xref)
        if pix.n > 4:   # CMYK等轉RGB
            pix = fitz.Pixmap(fitz.csRGB, pix)
        pix.save(f"img_p{page_index}_{xref}.png")

import camelot

# 自動模式（按頁號），嘗試 lattice 或 stream
tables = camelot.read_pdf("tables.pdf", pages="1-3", flavor="lattice")
print(tables.n)                 # 抽到了多少張表
df = tables[0].df               # 直接拿 pandas.DataFrame
tables.export("out.csv", f="csv", compress=True)

import tabula
dfs = tabula.read_pdf("tables.pdf", pages="all", multiple_tables=True)

ocrmypdf -l chi_sim+eng --rotate-pages --deskew input_scan.pdf searchable.pdf

from pdf2image import convert_from_path
import pytesseract

pages = convert_from_path("scan.pdf", dpi=300)  # 依賴 Poppler
full_text = []
for img in pages:
    txt = pytesseract.image_to_string(img, lang="chi_sim+eng")
    full_text.append(txt)
print("\n".join(full_text))

from tika import parser
parsed = parser.from_file("input.pdf")
print(parsed["content"])     # 純文本
print(parsed["metadata"])    # 元數(shù)據(jù)

import pypdfium2 as pdfium

pdf = pdfium.PdfDocument("input.pdf")
page = pdf.get_page(0)
textpage = page.get_textpage()
# 搜索關鍵字，返回范圍迭代器
for match in textpage.search("發(fā)票", match_case=False):
    # 獲取該命中范圍的包圍盒（可高亮/裁剪區(qū)域抽?。?
    rect = textpage.get_rect(match)
    print(rect.left, rect.top, rect.right, rect.bottom)

from pypdf import PdfReader

reader = PdfReader("has_attachments.pdf")
for name, blobs in reader.attachments.items():
    for i, content in enumerate(blobs):
        with open(f"{name}-{i}", "wb") as f:
            f.write(content)

# pypdf：讀表單域與值
from pypdf import PdfReader
reader = PdfReader("form.pdf")
fields = reader.get_fields()        # 或 reader.get_form_text_fields()
print(fields)

# pypdf 讀取注釋類型與位置
from pypdf import PdfReader
r = PdfReader("annotated.pdf")
for page in r.pages:
    if "/Annots" in page:
        for a in page["/Annots"]:
            obj = a.get_object()
            print(obj["/Subtype"], obj["/Rect"])

import pikepdf
pdf = pikepdf.open("input.pdf")
print(pdf.docinfo)            # 舊式 DocumentInfo（PDF 2.0 已廢棄但仍常見）
meta = pdf.open_metadata()    # XMP 元數(shù)據(jù)
print(meta)

import pdfplumber

with pdfplumber.open("input.pdf") as pdf:
    page = pdf.pages[0]
    # bbox = (x0, top, x1, bottom) ，單位：PDF points
    region = page.within_bbox((72, 72, 540, 200))
    print(region.extract_text())

"""
功能：
1) 先用 pypdf 試文本；太少/失敗 -> 判斷可能是掃描件 -> 走 OCRmyPDF 或 pdf2image+pytesseract
2) 可選：Camelot 抽表、PyMuPDF 抽圖片/鏈接/注釋、pypdf 抽附件，pikepdf 取元數(shù)據(jù)
3) 輸出 JSON：text/table/images/annotations/forms/attachments/metadata
"""
import json, os, subprocess, tempfile, shutil
from pypdf import PdfReader
import fitz

def is_text_pdf(path, min_chars=200):
    try:
        reader = PdfReader(path)
        s = "".join((p.extract_text() or "") for p in reader.pages[:5])
        return len(s.strip()) >= min_chars
    except Exception:
        return False

def ocr_if_needed(path):
    if is_text_pdf(path):
        return path  # 原樣返回
    # 嘗試用 OCRmyPDF（若未安裝，可改為 pdf2image+pytesseract）
    out = os.path.join(tempfile.gettempdir(), f"ocr_{os.path.basename(path)}")
    try:
        subprocess.run(
            ["ocrmypdf", "--skip-text", "-l", "chi_sim+eng", path, out],
            check=True, capture_output=True
        )
        return out
    except Exception:
        return path  # 回退：繼續(xù)用原文件（避免中斷）

def extract_all(path):
    path = ocr_if_needed(path)
    result = {"text": "", "tables": [], "images": [], "links": [], "annots": [],
              "attachments": [], "metadata": {}, "forms": {}}

    # 1. 文本（pypdf）
    r = PdfReader(path)
    result["text"] = "\n".join((p.extract_text() or "") for p in r.pages)

    # 2. 附件與表單
    try:
        result["attachments"] = list(r.attachments.keys())
    except Exception:
        pass
    try:
        result["forms"] = r.get_fields() or {}
    except Exception:
        pass

    # 3. 元數(shù)據(jù)（pikepdf 可更全面，這里用 pypdf 的 DocInfo 兜底）
    try:
        result["metadata"] = dict(r.metadata or {})
    except Exception:
        pass

    # 4. 圖片/鏈接/注釋（PyMuPDF）
    doc = fitz.open(path)
    for i, page in enumerate(doc):
        # 圖片
        for img in page.get_images(full=True):
            result["images"].append({"page": i+1, "xref": img[0], "width": img[2], "height": img[3]})
        # 鏈接
        for lk in page.get_links():
            result["links"].append({"page": i+1, **lk})
        # 注釋
        for a in page.annots() or []:
            result["annots"].append({"page": i+1, **(a.info or {})})

    return result

if __name__ == "__main__":
    data = extract_all("example.pdf")
    print(json.dumps(data, ensure_ascii=False, indent=2))