PyMuPDF与ReportLab坐标不兼容，绘制文本块位置异常求助

我尝试用ReportLab绘制PyMuPDF（fitz）识别出的文本块，但生成的PDF中文本块位置完全不对，以下是我的代码：

doc = fitz.open("demo.pdf")

canvas = Canvas("demo_.pdf", bottomup = True)


def draw_auto_fit_text_block(canvas, x_1, y_1, text_block_width, text_block_height, font_name, font_size, text_content):

    text_block_frame = Frame(x_1, y_1, text_block_width, text_block_height, topPadding = 0, leftPadding = 0, rightPadding = 0, bottomPadding = 0, showBoundary = 1)
    text_block_styles = ParagraphStyle(name = "Normal", fontName = font_name, fontSize = font_size)
    text_block_content = text_content.replace('\n','<br />\n')

    text_block_story = [Paragraph(text_block_content, style = text_block_styles)]
    text_block_story_inframe = KeepInFrame(text_block_width, text_block_height, text_block_story)
    text_block_frame.addFromList([text_block_story_inframe], canvas)

for page in doc:

    page_width = page.rect.width
    page_height = page.rect.height

    print("[page width]", page_width)
    print("[page height]", page_height)

    canvas.setPageSize((page_width, page_height))


    blocks = page.get_text("blocks")
    
    for block in blocks:

        block_content = block[4].replace("\n", " ").replace("- ", "-").strip()

        block_x_0 = block[0]
        block_y_0 = block[1]
        block_x_1 = block[2]
        block_y_1 = block[3]

        block_width = block_x_1 - block_x_0
        block_height = block_y_1 - block_y_0
        
        block_y_0 = page_height - block_y_0
        block_y_1 = page_height - block_y_0


        draw_auto_fit_text_block(canvas, block_x_0, block_y_0, block_width, block_height, font_name = "NimbusRomNo9L-Regu", font_size = 9.0, text_content = block_content)

    canvas.showPage()

canvas.save()

效果对比 #

• 生成的PDF效果：
• 原PDF效果：

1. 修复坐标转换逻辑 #

PyMuPDF和ReportLab的坐标系原点方向不同：

• PyMuPDF：左上角为原点(0,0)，y轴向下递增
• ReportLab（bottomup=True）：左下角为原点(0,0)，y轴向上递增

原代码中坐标转换完全错误，正确的转换应该是：

• 文本块在ReportLab中的左下角y坐标 = 页面高度 - PyMuPDF中块的底部y坐标（block_y_1）
• 块的高度保持block_y_1 - block_y_0不变

2. 保留文本换行格式 #

原代码把文本中的换行符替换成空格，导致段落结构丢失，应该保留换行并转换为ReportLab支持的<br/>标签。

3. 修正后的完整代码 #

from reportlab.pdfgen import Canvas
from reportlab.lib.styles import ParagraphStyle
from reportlab.platypus import Paragraph, Frame, KeepInFrame
import fitz

doc = fitz.open("demo.pdf")
canvas = Canvas("demo_.pdf", bottomup=True)

def draw_auto_fit_text_block(canvas, x, y, width, height, font_name, font_size, text_content):
    # 创建文本框，边界用于调试（可删除showBoundary=1）
    text_frame = Frame(x, y, width, height, 
                      topPadding=0, leftPadding=0, 
                      rightPadding=0, bottomPadding=0, 
                      showBoundary=1)
    style = ParagraphStyle(name="Normal", fontName=font_name, fontSize=font_size)
    # 把换行转成ReportLab支持的换行标签
    formatted_text = text_content.replace('\n', '<br />\n')
    paragraph = Paragraph(formatted_text, style)
    # 用KeepInFrame确保文本适应框大小，mode=shrink会自动缩小字体
    story = KeepInFrame(width, height, [paragraph], mode="shrink")
    text_frame.addFromList([story], canvas)

for page in doc:
    page_width = page.rect.width
    page_height = page.rect.height
    canvas.setPageSize((page_width, page_height))
    
    blocks = page.get_text("blocks")
    for block in blocks:
        block_content = block[4].replace("- ", "-").strip()  # 保留换行，只处理连字符
        x0, y0, x1, y1 = block[:4]
        block_width = x1 - x0
        block_height = y1 - y0
        
        # 正确转换坐标：ReportLab的左下角y = 页面高度 - 原块的底部y
        reportlab_y = page_height - y1
        
        draw_auto_fit_text_block(canvas, x0, reportlab_y, block_width, block_height,
                                font_name="NimbusRomNo9L-Regu", font_size=9.0,
                                text_content=block_content)
    canvas.showPage()

canvas.save()

额外说明 #

• 如果字体显示异常，确保ReportLab能找到指定字体，可通过reportlab.pdfbase.pdfmetrics.registerFont注册自定义字体
• showBoundary=1会显示文本块的边框，调试完成后可以删除该参数隐藏边框

内容的提问来源于stack exchange，提问作者CAO RUI