Python PyMuPDF fitz 教學：從pdf中抓文字、抓 fonts、抓表格; pip install PyMuPDF ; import fitz ; text_dict = page.get_text(“dict”) #type(page) is pymupdf.Page ; blocks:list[dict] = text_dict[‘blocks’] ; page.find_tables().tables [0].extract() ;如何判斷粗體字?

page.insert_text((x, y), span_text, fontsize=size, fontname=font)
x += fitz.get_text_length(span_text, fontname=font, fontsize=size)

page.insert_text((x, y), span_text, fontsize=size, fontname=font)
x += fitz.get_text_length(span_text, fontname=font, fontsize=size)

import os
import fitz

output_path = r"D:\Temp\fitz_multi_span_line.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

doc = fitz.open()
page = doc.new_page(width=595, height=842)

y = 100
x = 50

spans = [
    {"text": "Mixed font: ", "font": "Helvetica", "size": 12},
    {"text": "BOLD_PART", "font": "Helvetica-Bold", "size": 12},
    {"text": " normal tail", "font": "Helvetica", "size": 12},
]

for span in spans:
    page.insert_text(
        (x, y),
        span["text"],
        fontsize=span["size"],
        fontname=span["font"],
    )

    x += fitz.get_text_length(
        span["text"],
        fontname=span["font"],
        fontsize=span["size"],
    )

doc.save(output_path)
doc.close()

print(output_path)

import os
import fitz

output_path = r"D:\Temp\fitz_multi_span_line.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

doc = fitz.open()
page = doc.new_page(width=595, height=842)

y = 100
x = 50

spans = [
    {"text": "Mixed font: ", "font": "Helvetica", "size": 12},
    {"text": "BOLD_PART", "font": "Helvetica-Bold", "size": 12},
    {"text": " normal tail", "font": "Helvetica", "size": 12},
]

for span in spans:
    page.insert_text(
        (x, y),
        span["text"],
        fontsize=span["size"],
        fontname=span["font"],
    )

    x += fitz.get_text_length(
        span["text"],
        fontname=span["font"],
        fontsize=span["size"],
    )

doc.save(output_path)
doc.close()

print(output_path)

  line_spacing = 1.2  # 常見的標準行距倍率
  line_height = spans[0]["size"] * line_spacing

  y += line_height  # 往下移動到下一行

  line_spacing = 1.2  # 常見的標準行距倍率
  line_height = spans[0]["size"] * line_spacing

  y += line_height  # 往下移動到下一行

# %%
import os
import fitz

# 準備存檔路徑
output_path = r"D:\Temp\fitz_two_lines_demo.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

# 建立新 PDF 與頁面
doc = fitz.open()
page = doc.new_page(width=595, height=842)

y = 100
x = 50

# --- 繪製第一行 ---
span_line1 = {"text": "This is the first line.", "font": "Helvetica", "size": 18}
page.insert_text(
    (x, y),
    span_line1["text"],
    fontsize=span_line1["size"],
    fontname=span_line1["font"],
)

# --- 計算行高並換行 ---
line_spacing = 1.2  # 常見的標準行距倍率
line_height = span_line1["size"] * line_spacing

y += line_height   # 往下移動到下一行的 y 基準線
# x = 50           # 如果需要退回行首，可以將 x 重設 (這邊因為本身 x 沒改變，所以仍是 50)

# --- 繪製第二行 ---
span_line2 = {"text": "This is the second line, perfectly spaced.", "font": "Helvetica", "size": 18}
page.insert_text(
    (x, y),
    span_line2["text"],
    fontsize=span_line2["size"],
    fontname=span_line2["font"],
)

# 存檔
doc.save(output_path)
doc.close()

print("PDF 已經成功輸出至:", output_path)

# %%
import os
import fitz

# 準備存檔路徑
output_path = r"D:\Temp\fitz_two_lines_demo.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

# 建立新 PDF 與頁面
doc = fitz.open()
page = doc.new_page(width=595, height=842)

y = 100
x = 50

# --- 繪製第一行 ---
span_line1 = {"text": "This is the first line.", "font": "Helvetica", "size": 18}
page.insert_text(
    (x, y),
    span_line1["text"],
    fontsize=span_line1["size"],
    fontname=span_line1["font"],
)

# --- 計算行高並換行 ---
line_spacing = 1.2  # 常見的標準行距倍率
line_height = span_line1["size"] * line_spacing

y += line_height   # 往下移動到下一行的 y 基準線
# x = 50           # 如果需要退回行首，可以將 x 重設 (這邊因為本身 x 沒改變，所以仍是 50)

# --- 繪製第二行 ---
span_line2 = {"text": "This is the second line, perfectly spaced.", "font": "Helvetica", "size": 18}
page.insert_text(
    (x, y),
    span_line2["text"],
    fontsize=span_line2["size"],
    fontname=span_line2["font"],
)

# 存檔
doc.save(output_path)
doc.close()

print("PDF 已經成功輸出至:", output_path)

import os
import fitz

output_path = r"D:\Temp\fitz_write_demo.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

doc = fitz.open()
page = doc.new_page(width=595, height=842)

page.insert_text(
    (50, 60),
    "PyMuPDF Demo Title",
    fontsize=18,
    fontname="Helvetica-Bold",
)

page.insert_text(
    (50, 95),
    "This is a normal line.",
    fontsize=11,
    fontname="Times-Roman",
)

page.insert_text(
    (50, 125),
    "Mixed font: ",
    fontsize=12,
    fontname="Helvetica",
)
page.insert_text(
    (120, 125),
    "BOLD_PART",
    fontsize=12,
    fontname="Helvetica-Bold",
)
page.insert_text(
    (195, 125),
    " normal tail",
    fontsize=12,
    fontname="Helvetica",
)

doc.save(output_path)
doc.close()

print(output_path)

import os
import fitz

output_path = r"D:\Temp\fitz_write_demo.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

doc = fitz.open()
page = doc.new_page(width=595, height=842)

page.insert_text(
    (50, 60),
    "PyMuPDF Demo Title",
    fontsize=18,
    fontname="Helvetica-Bold",
)

page.insert_text(
    (50, 95),
    "This is a normal line.",
    fontsize=11,
    fontname="Times-Roman",
)

page.insert_text(
    (50, 125),
    "Mixed font: ",
    fontsize=12,
    fontname="Helvetica",
)
page.insert_text(
    (120, 125),
    "BOLD_PART",
    fontsize=12,
    fontname="Helvetica-Bold",
)
page.insert_text(
    (195, 125),
    " normal tail",
    fontsize=12,
    fontname="Helvetica",
)

doc.save(output_path)
doc.close()

print(output_path)

import fitz

pdf_path = r"D:\Temp\fitz_write_demo.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

print(page.get_text())
doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_write_demo.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

print(page.get_text())
doc.close()

import os
import fitz

output_path = r"D:\Temp\fitz_write_table_demo.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

doc = fitz.open()
page = doc.new_page(width=595, height=842)

left = 50
right = 430
top = 120
row_h = 28
col_x = [left, 180, 300, right]
rows = [top + i * row_h for i in range(5)]

for x in col_x:
    page.draw_line((x, rows[0]), (x, rows[-1]), width=1)

for y in rows:
    page.draw_line((left, y), (right, y), width=1)

headers = ["Item", "Class", "Capacity"]
for idx, text in enumerate(headers):
    page.insert_text(
        (col_x[idx] + 8, top + 18),
        text,
        fontsize=11,
        fontname="Helvetica-Bold",
    )

data = [
    ("M", "Module", "16GB"),
    ("R", "RDIMM", "32GB"),
    ("U", "UDIMM", "64GB"),
]

for row_idx, row in enumerate(data, start=1):
    y = top + row_idx * row_h + 18
    for col_idx, value in enumerate(row):
        page.insert_text(
            (col_x[col_idx] + 8, y),
            value,
            fontsize=11,
            fontname="Helvetica",
        )

doc.save(output_path)
doc.close()

print(output_path)

import os
import fitz

output_path = r"D:\Temp\fitz_write_table_demo.pdf"
os.makedirs(os.path.dirname(output_path), exist_ok=True)

doc = fitz.open()
page = doc.new_page(width=595, height=842)

left = 50
right = 430
top = 120
row_h = 28
col_x = [left, 180, 300, right]
rows = [top + i * row_h for i in range(5)]

for x in col_x:
    page.draw_line((x, rows[0]), (x, rows[-1]), width=1)

for y in rows:
    page.draw_line((left, y), (right, y), width=1)

headers = ["Item", "Class", "Capacity"]
for idx, text in enumerate(headers):
    page.insert_text(
        (col_x[idx] + 8, top + 18),
        text,
        fontsize=11,
        fontname="Helvetica-Bold",
    )

data = [
    ("M", "Module", "16GB"),
    ("R", "RDIMM", "32GB"),
    ("U", "UDIMM", "64GB"),
]

for row_idx, row in enumerate(data, start=1):
    y = top + row_idx * row_h + 18
    for col_idx, value in enumerate(row):
        page.insert_text(
            (col_x[col_idx] + 8, y),
            value,
            fontsize=11,
            fontname="Helvetica",
        )

doc.save(output_path)
doc.close()

print(output_path)

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)

print(len(doc))
doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)

print(len(doc))
doc.close()

page
  -> blocks
    -> lines
      -> spans

page
  -> blocks
    -> lines
      -> spans

import fitz
from typing import TypeAlias

BlockTuple: TypeAlias = tuple[float, float, float, float, str, int, int]
# (x0, y0, x1, y1, text, block_no, block_type)
pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

blocks: list[BlockTuple] = page.get_text("blocks")
for block in blocks[:5]:
    print(block)

doc.close()

import fitz
from typing import TypeAlias

BlockTuple: TypeAlias = tuple[float, float, float, float, str, int, int]
# (x0, y0, x1, y1, text, block_no, block_type)
pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

blocks: list[BlockTuple] = page.get_text("blocks")
for block in blocks[:5]:
    print(block)

doc.close()

tuple[float, float, float, float, str, int, int]

tuple[float, float, float, float, str, int, int]

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

text_dict = page.get_text("dict")
print(text_dict.keys())

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

text_dict = page.get_text("dict")
print(text_dict.keys())

doc.close()

text_dict.keys()
# dict_keys(['width', 'height', 'blocks'])

text_dict.keys()
# dict_keys(['width', 'height', 'blocks'])

text_dict = page.get_text("dict") #type(page) is pymupdf.Page
  -> blocks :list[dict]
    -> block :dict #dict_keys(['number', 'type', 'bbox', 'lines'])
    
      -> lines :list[dict]
        -> line :dict #dict_keys(['spans', 'wmode', 'dir', 'bbox'])
        
          -> spans :list[dict]
            -> span :dict #dict_keys(['size', 'flags', 'bidi', 'char_flags', 'font', 'color', 'alpha', 'ascender', 'descender', 'text', 'origin', 'bbox'])

text_dict = page.get_text("dict") #type(page) is pymupdf.Page
  -> blocks :list[dict]
    -> block :dict #dict_keys(['number', 'type', 'bbox', 'lines'])
    
      -> lines :list[dict]
        -> line :dict #dict_keys(['spans', 'wmode', 'dir', 'bbox'])
        
          -> spans :list[dict]
            -> span :dict #dict_keys(['size', 'flags', 'bidi', 'char_flags', 'font', 'color', 'alpha', 'ascender', 'descender', 'text', 'origin', 'bbox'])

import fitz
from typing import Any

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)

for i, page in enumerate(doc, start=1):
    # 1. 取得整頁的 dictionary
    text_dict: dict[str, Any] = page.get_text("dict")

    # 2. 取得所有的 blocks (這裡包含文字、圖片等各種區塊)
    all_blocks: list[dict[str, Any]] = text_dict.get("blocks", [])

    # 3. 過濾出純文字的 blocks (type == 0)
    text_blocks: list[dict[str, Any]] = [
        block for block in all_blocks if block.get("type") == 0
    ]

    # 4. 展開所有的 lines
    text_lines: list[dict[str, Any]] = []
    for block in text_blocks:
        lines: list[dict[str, Any]] = block.get("lines", [])
        text_lines.extend(lines)

    # 5. 計算總數
    line_count: int = len(text_lines)

    # 計算所有 lines 裡面的 spans 總數
    span_count: int = 0
    for line in text_lines:
        spans: list[dict[str, Any]] = line.get("spans", [])
        span_count += len(spans)

    print(f"page={i} text_blocks={len(text_blocks)} lines={line_count} spans={span_count}")

doc.close()

import fitz
from typing import Any

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)

for i, page in enumerate(doc, start=1):
    # 1. 取得整頁的 dictionary
    text_dict: dict[str, Any] = page.get_text("dict")

    # 2. 取得所有的 blocks (這裡包含文字、圖片等各種區塊)
    all_blocks: list[dict[str, Any]] = text_dict.get("blocks", [])

    # 3. 過濾出純文字的 blocks (type == 0)
    text_blocks: list[dict[str, Any]] = [
        block for block in all_blocks if block.get("type") == 0
    ]

    # 4. 展開所有的 lines
    text_lines: list[dict[str, Any]] = []
    for block in text_blocks:
        lines: list[dict[str, Any]] = block.get("lines", [])
        text_lines.extend(lines)

    # 5. 計算總數
    line_count: int = len(text_lines)

    # 計算所有 lines 裡面的 spans 總數
    span_count: int = 0
    for line in text_lines:
        spans: list[dict[str, Any]] = line.get("spans", [])
        span_count += len(spans)

    print(f"page={i} text_blocks={len(text_blocks)} lines={line_count} spans={span_count}")

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page:pymupdf.Page = doc[0]

text_dict = page.get_text("dict")
print("text_dict.keys() =", text_dict.keys())

blocks = text_dict.get("blocks", [])
if blocks:
    first_block = blocks[0]
    print("first_block.keys() =", first_block.keys())

    lines = first_block.get("lines", [])
    if lines:
        first_line = lines[0]
        print("first_line.keys() =", first_line.keys())

        spans = first_line.get("spans", [])
        if spans:
            first_span = spans[0]
            print("first_span.keys() =", first_span.keys())

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page:pymupdf.Page = doc[0]

text_dict = page.get_text("dict")
print("text_dict.keys() =", text_dict.keys())

blocks = text_dict.get("blocks", [])
if blocks:
    first_block = blocks[0]
    print("first_block.keys() =", first_block.keys())

    lines = first_block.get("lines", [])
    if lines:
        first_line = lines[0]
        print("first_line.keys() =", first_line.keys())

        spans = first_line.get("spans", [])
        if spans:
            first_span = spans[0]
            print("first_span.keys() =", first_span.keys())

doc.close()

text_dict.keys() = dict_keys(['width', 'height', 'blocks'])
first_block.keys() = dict_keys(['number', 'type', 'bbox', 'lines'])
first_line.keys() = dict_keys(['spans', 'wmode', 'dir', 'bbox'])
first_span.keys() = dict_keys(['size', 'flags', 'bidi', 'char_flags', 
'font', 'color', 'alpha', 'ascender', 'descender', 
'text', 'origin', 'bbox'])

text_dict.keys() = dict_keys(['width', 'height', 'blocks'])
first_block.keys() = dict_keys(['number', 'type', 'bbox', 'lines'])
first_line.keys() = dict_keys(['spans', 'wmode', 'dir', 'bbox'])
first_span.keys() = dict_keys(['size', 'flags', 'bidi', 'char_flags', 
'font', 'color', 'alpha', 'ascender', 'descender', 
'text', 'origin', 'bbox'])

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]
text_dict = page.get_text("dict")

for block in text_dict.get("blocks", []):
    if block.get("type") != 0:
        continue

    for line in block.get("lines", []):
        spans = line.get("spans", [])
        if not spans:
            continue

        line_text = "".join(span.get("text", "") for span in spans).strip()
        if not line_text:
            continue

        print(f"line_text={line_text!r}, span_count={len(spans)}")
        for idx, span in enumerate(spans, start=1):
            print(
                f"  span{idx}: text={span.get('text', '')!r}, "
                f"font={span.get('font', '')!r}, size={span.get('size', 0)}"
            )

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]
text_dict = page.get_text("dict")

for block in text_dict.get("blocks", []):
    if block.get("type") != 0:
        continue

    for line in block.get("lines", []):
        spans = line.get("spans", [])
        if not spans:
            continue

        line_text = "".join(span.get("text", "") for span in spans).strip()
        if not line_text:
            continue

        print(f"line_text={line_text!r}, span_count={len(spans)}")
        for idx, span in enumerate(spans, start=1):
            print(
                f"  span{idx}: text={span.get('text', '')!r}, "
                f"font={span.get('font', '')!r}, size={span.get('size', 0)}"
            )

doc.close()

first_span.keys() = dict_keys(['size', 'flags', 'bidi', 'char_flags', 
'font', 'color', 'alpha', 'ascender', 'descender', 
'text', 'origin', 'bbox'])

first_span.keys() = dict_keys(['size', 'flags', 'bidi', 'char_flags', 
'font', 'color', 'alpha', 'ascender', 'descender', 
'text', 'origin', 'bbox'])

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

text_dict = page.get_text("dict")

for block in text_dict.get("blocks", []):
    if block.get("type") != 0:
        continue

    for line in block.get("lines", []):
        for span in line.get("spans", []):
            print({
                "text": span.get("text", ""),
                "font": span.get("font", ""),
                "size": span.get("size", 0),
                "flags": span.get("flags", 0),
                "color": span.get("color", ""),
                "bbox": span.get("bbox", None),
            })

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

text_dict = page.get_text("dict")

for block in text_dict.get("blocks", []):
    if block.get("type") != 0:
        continue

    for line in block.get("lines", []):
        for span in line.get("spans", []):
            print({
                "text": span.get("text", ""),
                "font": span.get("font", ""),
                "size": span.get("size", 0),
                "flags": span.get("flags", 0),
                "color": span.get("color", ""),
                "bbox": span.get("bbox", None),
            })

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]
text_dict:dict = page.get_text("dict")
# dict_keys(['width', 'height', 'blocks'])

for block in text_dict.get("blocks", []):
    # block: dict
    # dict_keys(['number', 'type', 'bbox', 'lines'])
    if block.get("type") != 0:
        continue

    for line in block.get("lines", []):
        # line: dict
        # dict_keys(['spans', 'wmode', 'dir', 'bbox'])
        spans = line.get("spans", [])
        # spans:list[dict]
        if len(spans) >= 2:
            print("line_text =", "".join(span.get("text", "") for span in spans))
            for span in spans:
                #span.keys()
                #dict_keys(['size', 'flags', 'bidi', 'char_flags', 'font', 'color', 'alpha', 'ascender', 'descender', 'text', 'origin', 'bbox'])
                print({
                    "text": span.get("text", ""),
                    "font": span.get("font", ""),
                    "size": span.get("size", 0),
                    "flags": span.get("flags", 0),
                })

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]
text_dict:dict = page.get_text("dict")
# dict_keys(['width', 'height', 'blocks'])

for block in text_dict.get("blocks", []):
    # block: dict
    # dict_keys(['number', 'type', 'bbox', 'lines'])
    if block.get("type") != 0:
        continue

    for line in block.get("lines", []):
        # line: dict
        # dict_keys(['spans', 'wmode', 'dir', 'bbox'])
        spans = line.get("spans", [])
        # spans:list[dict]
        if len(spans) >= 2:
            print("line_text =", "".join(span.get("text", "") for span in spans))
            for span in spans:
                #span.keys()
                #dict_keys(['size', 'flags', 'bidi', 'char_flags', 'font', 'color', 'alpha', 'ascender', 'descender', 'text', 'origin', 'bbox'])
                print({
                    "text": span.get("text", ""),
                    "font": span.get("font", ""),
                    "size": span.get("size", 0),
                    "flags": span.get("flags", 0),
                })

doc.close()

# %%
import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
rows = []

for page_no, page in enumerate(doc, start=1):
    text_dict = page.get_text("dict")

    for block_no, block in enumerate(text_dict.get("blocks", [])):
        if block.get("type") != 0:
            continue

        for line_no, line in enumerate(block.get("lines", [])):
            spans = line.get("spans", [])
            if not spans:
                continue

            line_text = "".join(span.get("text", "") for span in spans)
            if not line_text.strip():
                continue

            main_span = max(spans, key=lambda s: len(s.get("text", "")))
            span_texts = [span.get("text", "") for span in spans]
            span_fonts = [span.get("font", "") for span in spans]
            unique_fonts = sorted(set(span_fonts))
            font_char_counts = {}
            for span in spans:
                font_name = span.get("font", "")
                font_char_counts[font_name] = font_char_counts.get(font_name, 0) + len(span.get("text", ""))
            fonts_summary = ", ".join(
                f"{font_name}({char_count} chars)"
                for font_name, char_count in font_char_counts.items()
            )

            rows.append({
                "page": page_no,
                "block_no": block_no,
                "line_no": line_no,
                "text": line_text,
                "font_main": main_span.get("font", ""),
                "size_main": main_span.get("size", 0),
                "flags_main": main_span.get("flags", 0),
                "span_count": len(spans),
                "span_texts": span_texts,
                "span_fonts": span_fonts,
                "font_count": len(unique_fonts),
                "has_mixed_fonts": len(unique_fonts) >= 2,
                "fonts_summary": fonts_summary,
            })

doc.close()

df = pd.DataFrame(rows)
df[df["span_count"] >= 2][
    [
        "page",
        "block_no",
        "line_no",
        "text",
        "span_count",
        "fonts_summary",
        "span_texts",
        "span_fonts",
        "has_mixed_fonts",
    ]
].head(10)

# %%
import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
rows = []

for page_no, page in enumerate(doc, start=1):
    text_dict = page.get_text("dict")

    for block_no, block in enumerate(text_dict.get("blocks", [])):
        if block.get("type") != 0:
            continue

        for line_no, line in enumerate(block.get("lines", [])):
            spans = line.get("spans", [])
            if not spans:
                continue

            line_text = "".join(span.get("text", "") for span in spans)
            if not line_text.strip():
                continue

            main_span = max(spans, key=lambda s: len(s.get("text", "")))
            span_texts = [span.get("text", "") for span in spans]
            span_fonts = [span.get("font", "") for span in spans]
            unique_fonts = sorted(set(span_fonts))
            font_char_counts = {}
            for span in spans:
                font_name = span.get("font", "")
                font_char_counts[font_name] = font_char_counts.get(font_name, 0) + len(span.get("text", ""))
            fonts_summary = ", ".join(
                f"{font_name}({char_count} chars)"
                for font_name, char_count in font_char_counts.items()
            )

            rows.append({
                "page": page_no,
                "block_no": block_no,
                "line_no": line_no,
                "text": line_text,
                "font_main": main_span.get("font", ""),
                "size_main": main_span.get("size", 0),
                "flags_main": main_span.get("flags", 0),
                "span_count": len(spans),
                "span_texts": span_texts,
                "span_fonts": span_fonts,
                "font_count": len(unique_fonts),
                "has_mixed_fonts": len(unique_fonts) >= 2,
                "fonts_summary": fonts_summary,
            })

doc.close()

df = pd.DataFrame(rows)
df[df["span_count"] >= 2][
    [
        "page",
        "block_no",
        "line_no",
        "text",
        "span_count",
        "fonts_summary",
        "span_texts",
        "span_fonts",
        "has_mixed_fonts",
    ]
].head(10)

import fitz
import pandas as pd
from typing import Any, Iterator


def iter_text_lines(doc: fitz.Document) -> Iterator[tuple[int, int, int, dict[str, Any]]]:
    for page_no, page in enumerate(doc, start=1):
        text_dict: dict[str, Any] = page.get_text("dict")

        for block_no, block in enumerate(text_dict.get("blocks", [])):
            if block.get("type") != 0:
                continue

            for line_no, line in enumerate(block.get("lines", [])):
                yield page_no, block_no, line_no, line


pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
rows = []

for page_no, block_no, line_no, line in iter_text_lines(doc):
    spans = line.get("spans", [])
    if not spans:
        continue

    line_text = "".join((span.get("text") or "") for span in spans)
    if not line_text.strip():
        continue

    main_span = max(spans, key=lambda s: len(s.get("text") or ""))
    span_texts = [(span.get("text") or "") for span in spans]
    span_fonts = [(span.get("font") or "") for span in spans]
    unique_fonts = sorted(set(span_fonts))

    font_char_counts: dict[str, int] = {}
    for span in spans:
        font_name = span.get("font") or "<UNKNOWN_FONT>"
        span_text = span.get("text") or ""
        font_char_counts[font_name] = font_char_counts.get(font_name, 0) + len(span_text)

    fonts_summary = ", ".join(
        f"{font_name}({char_count} chars)"
        for font_name, char_count in font_char_counts.items()
    )

    rows.append({
        "page": page_no,
        "block_no": block_no,
        "line_no": line_no,
        "text": line_text,
        "font_main": main_span.get("font") or "",
        "size_main": main_span.get("size", 0),
        "flags_main": main_span.get("flags", 0),
        "span_count": len(spans),
        "span_texts": span_texts,
        "span_fonts": span_fonts,
        "font_count": len(unique_fonts),
        "has_mixed_fonts": len(unique_fonts) >= 2,
        "fonts_summary": fonts_summary,
    })

doc.close()

df = pd.DataFrame(rows)
df[df["span_count"] >= 2][
    [
        "page",
        "block_no",
        "line_no",
        "text",
        "span_count",
        "fonts_summary",
        "span_texts",
        "span_fonts",
        "has_mixed_fonts",
    ]
].head(10)

import fitz
import pandas as pd
from typing import Any, Iterator


def iter_text_lines(doc: fitz.Document) -> Iterator[tuple[int, int, int, dict[str, Any]]]:
    for page_no, page in enumerate(doc, start=1):
        text_dict: dict[str, Any] = page.get_text("dict")

        for block_no, block in enumerate(text_dict.get("blocks", [])):
            if block.get("type") != 0:
                continue

            for line_no, line in enumerate(block.get("lines", [])):
                yield page_no, block_no, line_no, line


pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
rows = []

for page_no, block_no, line_no, line in iter_text_lines(doc):
    spans = line.get("spans", [])
    if not spans:
        continue

    line_text = "".join((span.get("text") or "") for span in spans)
    if not line_text.strip():
        continue

    main_span = max(spans, key=lambda s: len(s.get("text") or ""))
    span_texts = [(span.get("text") or "") for span in spans]
    span_fonts = [(span.get("font") or "") for span in spans]
    unique_fonts = sorted(set(span_fonts))

    font_char_counts: dict[str, int] = {}
    for span in spans:
        font_name = span.get("font") or "<UNKNOWN_FONT>"
        span_text = span.get("text") or ""
        font_char_counts[font_name] = font_char_counts.get(font_name, 0) + len(span_text)

    fonts_summary = ", ".join(
        f"{font_name}({char_count} chars)"
        for font_name, char_count in font_char_counts.items()
    )

    rows.append({
        "page": page_no,
        "block_no": block_no,
        "line_no": line_no,
        "text": line_text,
        "font_main": main_span.get("font") or "",
        "size_main": main_span.get("size", 0),
        "flags_main": main_span.get("flags", 0),
        "span_count": len(spans),
        "span_texts": span_texts,
        "span_fonts": span_fonts,
        "font_count": len(unique_fonts),
        "has_mixed_fonts": len(unique_fonts) >= 2,
        "fonts_summary": fonts_summary,
    })

doc.close()

df = pd.DataFrame(rows)
df[df["span_count"] >= 2][
    [
        "page",
        "block_no",
        "line_no",
        "text",
        "span_count",
        "fonts_summary",
        "span_texts",
        "span_fonts",
        "has_mixed_fonts",
    ]
].head(10)

def is_bold_span(font_name: str, flags: int) -> bool:
    lower_font_name = font_name.lower()
    if "bold" in lower_font_name:
        return True

    return bool(flags & 16)

def is_bold_span(font_name: str, flags: int) -> bool:
    lower_font_name = font_name.lower()
    if "bold" in lower_font_name:
        return True

    return bool(flags & 16)

print(16 & 16)         # 16
print(bool(16 & 16))   # True

print(17 & 16)         # 16
print(bool(17 & 16))   # True

print(4 & 16)          # 0
print(bool(4 & 16))    # False

print(16 & 16)         # 16
print(bool(16 & 16))   # True

print(17 & 16)         # 16
print(bool(17 & 16))   # True

print(4 & 16)          # 0
print(bool(4 & 16))    # False

page.find_tables()

page.find_tables()

page.find_tables()
  -> TableFinder
      -> .tables
          -> [Table, Table, ...]

page.find_tables()
  -> TableFinder
      -> .tables
          -> [Table, Table, ...]

tables_result = page.find_tables()   # TableFinder
tables = tables_result.tables        # list[Table]
first_table = tables[0]              # Table
rows = first_table.extract()         # list[list[str]]

tables_result = page.find_tables()   # TableFinder
tables = tables_result.tables        # list[Table]
first_table = tables[0]              # Table
rows = first_table.extract()         # list[list[str]]

rows
    -> [row0, row1, row2, ...]
             -> [cell0, cell1, cell2, ...]

rows
    -> [row0, row1, row2, ...]
             -> [cell0, cell1, cell2, ...]

[
        ['Item', 'Class', 'Capacity'],
        ['M', 'Module', '16GB'],
        ['R', 'RDIMM', '32GB'],
        ['U', 'UDIMM', '64GB'],
]

[
        ['Item', 'Class', 'Capacity'],
        ['M', 'Module', '16GB'],
        ['R', 'RDIMM', '32GB'],
        ['U', 'UDIMM', '64GB'],
]

print(rows)
print(rows[0])      # header row
print(rows[1])      # first data row
print(rows[1][0])   # first cell of first data row

print(rows)
print(rows[0])      # header row
print(rows[1])      # first data row
print(rows[1][0])   # first cell of first data row

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

tables_result = page.find_tables()
tables = tables_result.tables

print(len(tables))
for table in tables:
    print(table.bbox)
    for row in table.extract():
        print(row)

doc.close()

import fitz

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

tables_result = page.find_tables()
tables = tables_result.tables

print(len(tables))
for table in tables:
    print(table.bbox)
    for row in table.extract():
        print(row)

doc.close()

import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

tables = page.find_tables().tables

if tables:
    table_df = pd.DataFrame(tables[0].extract())
    print(table_df)

doc.close()

import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

tables = page.find_tables().tables

if tables:
    table_df = pd.DataFrame(tables[0].extract())
    print(table_df)

doc.close()

import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

tables = page.find_tables().tables

if tables:
    rows = tables[0].extract()
    header = rows[0]
    body = rows[1:]
    table_df = pd.DataFrame(body, columns=header)
    print(table_df)

doc.close()

import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)
page = doc[0]

tables = page.find_tables().tables

if tables:
    rows = tables[0].extract()
    header = rows[0]
    body = rows[1:]
    table_df = pd.DataFrame(body, columns=header)
    print(table_df)

doc.close()

import pandas as pd

def normalize_table_df(table_df: pd.DataFrame) -> pd.DataFrame:
    normalized_df = table_df.copy()
    normalized_df = normalized_df.map(
        lambda value: value.strip() if isinstance(value, str) else value
    )
    normalized_df = normalized_df.replace("", pd.NA)
    normalized_df = normalized_df.dropna(axis=0, how="all")
    normalized_df = normalized_df.dropna(axis=1, how="all")
    return normalized_df.reset_index(drop=True)


def is_meaningful_table(table_df: pd.DataFrame) -> bool:
    if table_df.empty:
        return False

    # pandas DataFrame 的 .sum() 行為說明：
    # 1. table_df.notna()：把整個表格所有的格子變成 True/False (有值為 True, 沒值或 NA 為 False)
    # 2. 第一次 .sum()：會「逐個直行(column)」把 True 當成 1 算總和，回傳的是一個 Series，例如：colA 有 2 個值, colB 有 1 個值
    # 3. 第二次 .sum()：把剛剛算出來的 Series (每一行的總數) 再全部加總起來，變成一個單一的整數 (整張表有幾個非空值)
    non_empty_cells = int(table_df.notna().sum().sum())
    
    # 這裡的邏輯是：如果整張表格有值(非空)的格子加起來不到 3 個，我們就當它是誤判的垃圾碎片
    return non_empty_cells >= 3

import pandas as pd

def normalize_table_df(table_df: pd.DataFrame) -> pd.DataFrame:
    normalized_df = table_df.copy()
    normalized_df = normalized_df.map(
        lambda value: value.strip() if isinstance(value, str) else value
    )
    normalized_df = normalized_df.replace("", pd.NA)
    normalized_df = normalized_df.dropna(axis=0, how="all")
    normalized_df = normalized_df.dropna(axis=1, how="all")
    return normalized_df.reset_index(drop=True)


def is_meaningful_table(table_df: pd.DataFrame) -> bool:
    if table_df.empty:
        return False

    # pandas DataFrame 的 .sum() 行為說明：
    # 1. table_df.notna()：把整個表格所有的格子變成 True/False (有值為 True, 沒值或 NA 為 False)
    # 2. 第一次 .sum()：會「逐個直行(column)」把 True 當成 1 算總和，回傳的是一個 Series，例如：colA 有 2 個值, colB 有 1 個值
    # 3. 第二次 .sum()：把剛剛算出來的 Series (每一行的總數) 再全部加總起來，變成一個單一的整數 (整張表有幾個非空值)
    non_empty_cells = int(table_df.notna().sum().sum())
    
    # 這裡的邏輯是：如果整張表格有值(非空)的格子加起來不到 3 個，我們就當它是誤判的垃圾碎片
    return non_empty_cells >= 3

import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)

raw_rows = []
tables_index = []

for page_no, page in enumerate(doc, start=1):
    text_dict = page.get_text("dict")

    for block_no, block in enumerate(text_dict.get("blocks", [])):
        if block.get("type") != 0:
            continue

        for line_no, line in enumerate(block.get("lines", [])):
            spans = line.get("spans", [])
            if not spans:
                continue

            text = "".join(span.get("text", "") for span in spans)
            if not text.strip():
                continue

            main_span = max(spans, key=lambda s: len(s.get("text", "")))
            raw_rows.append({
                "page": page_no,
                "block_no": block_no,
                "line_no": line_no,
                "text": text,
                "font_main": main_span.get("font", ""),
                "flags_main": main_span.get("flags", 0),
            })

    tables = page.find_tables().tables
    for table_idx, table in enumerate(tables, start=1):
        table_df = pd.DataFrame(table.extract())
        tables_index.append({
            "page": page_no,
            "table_id": f"table{table_idx}",
            "bbox": table.bbox,
            "row_count": table_df.shape[0],
            "col_count": table_df.shape[1],
        })

doc.close()

raw_text_df = pd.DataFrame(raw_rows)
tables_index_df = pd.DataFrame(tables_index)

print(raw_text_df.head())
print(tables_index_df)

import fitz
import pandas as pd

pdf_path = r"D:\Temp\fitz_demo_tutorial.pdf"
doc = fitz.open(pdf_path)

raw_rows = []
tables_index = []

for page_no, page in enumerate(doc, start=1):
    text_dict = page.get_text("dict")

    for block_no, block in enumerate(text_dict.get("blocks", [])):
        if block.get("type") != 0:
            continue

        for line_no, line in enumerate(block.get("lines", [])):
            spans = line.get("spans", [])
            if not spans:
                continue

            text = "".join(span.get("text", "") for span in spans)
            if not text.strip():
                continue

            main_span = max(spans, key=lambda s: len(s.get("text", "")))
            raw_rows.append({
                "page": page_no,
                "block_no": block_no,
                "line_no": line_no,
                "text": text,
                "font_main": main_span.get("font", ""),
                "flags_main": main_span.get("flags", 0),
            })

    tables = page.find_tables().tables
    for table_idx, table in enumerate(tables, start=1):
        table_df = pd.DataFrame(table.extract())
        tables_index.append({
            "page": page_no,
            "table_id": f"table{table_idx}",
            "bbox": table.bbox,
            "row_count": table_df.shape[0],
            "col_count": table_df.shape[1],
        })

doc.close()

raw_text_df = pd.DataFrame(raw_rows)
tables_index_df = pd.DataFrame(tables_index)

print(raw_text_df.head())
print(tables_index_df)