PDFTranslate / pdf2zh /converter.py
sanbo
update sth. at 2024-11-26 16:15:47
9b0f4a0
raw
history blame
53.9 kB
from pdf2zh.utils import (
AnyIO,
Matrix,
PathSegment,
Point,
Rect,
apply_matrix_pt,
bbox2str,
enc,
make_compat_str,
mult_matrix,
matrix_scale,
)
from pdf2zh.pdftypes import PDFStream
from pdf2zh.pdfpage import PDFPage
from pdf2zh.pdfinterp import PDFGraphicState, PDFResourceManager
from pdf2zh.pdffont import PDFFont, PDFUnicodeNotDefined, PDFCIDFont
from pdf2zh.pdfexceptions import PDFValueError
from pdf2zh.pdfdevice import PDFTextDevice
from pdf2zh.pdfcolor import PDFColorSpace
from pdf2zh.layout import (
LAParams,
LTAnno,
LTChar,
LTComponent,
LTCurve,
LTFigure,
LTImage,
LTItem,
LTLayoutContainer,
LTLine,
LTPage,
LTRect,
LTText,
LTTextBox,
LTTextBoxVertical,
LTTextGroup,
LTTextLine,
TextGroupElement,
)
from pdf2zh.image import ImageWriter
from pdf2zh import utils
import io
import logging
import re
from typing import (
BinaryIO,
Dict,
Generic,
List,
Optional,
Sequence,
TextIO,
Tuple,
TypeVar,
Union,
cast,
)
import concurrent.futures
import numpy as np
import unicodedata
from tenacity import retry, wait_fixed
from pdf2zh import cache
from pdf2zh.translator import (
BaseTranslator,
GoogleTranslator,
DeepLTranslator,
DeepLXTranslator,
OllamaTranslator,
OpenAITranslator,
AzureTranslator,
TencentTranslator,
)
def remove_control_characters(s):
return "".join(ch for ch in s if unicodedata.category(ch)[0] != "C")
log = logging.getLogger(__name__)
class PDFLayoutAnalyzer(PDFTextDevice):
cur_item: LTLayoutContainer
ctm: Matrix
def __init__(
self,
rsrcmgr: PDFResourceManager,
pageno: int = 1,
laparams: Optional[LAParams] = None,
) -> None:
PDFTextDevice.__init__(self, rsrcmgr)
self.pageno = pageno
self.laparams = laparams
self._stack: List[LTLayoutContainer] = []
def begin_page(self, page: PDFPage, ctm: Matrix) -> None:
# (x0, y0, x1, y1) = page.mediabox
(x0, y0, x1, y1) = page.cropbox
(x0, y0) = apply_matrix_pt(ctm, (x0, y0))
(x1, y1) = apply_matrix_pt(ctm, (x1, y1))
mediabox = (0, 0, abs(x0 - x1), abs(y0 - y1))
self.cur_item = LTPage(page.pageno, mediabox)
def end_page(self, page: PDFPage):
assert not self._stack, str(len(self._stack))
assert isinstance(self.cur_item, LTPage), str(type(self.cur_item))
# 取消默认排版分析
# if self.laparams is not None:
# self.cur_item.analyze(self.laparams)
self.pageno += 1
return self.receive_layout(self.cur_item)
def begin_figure(self, name: str, bbox: Rect, matrix: Matrix) -> None:
self._stack.append(self.cur_item)
self.cur_item = LTFigure(name, bbox, mult_matrix(matrix, self.ctm))
self.cur_item.pageid = self._stack[-1].pageid
def end_figure(self, _: str) -> None:
fig = self.cur_item
assert isinstance(self.cur_item, LTFigure), str(type(self.cur_item))
self.cur_item = self._stack.pop()
self.cur_item.add(fig)
return self.receive_layout(fig)
def render_image(self, name: str, stream: PDFStream) -> None:
assert isinstance(self.cur_item, LTFigure), str(type(self.cur_item))
item = LTImage(
name,
stream,
(self.cur_item.x0, self.cur_item.y0, self.cur_item.x1, self.cur_item.y1),
)
self.cur_item.add(item)
def paint_path(
self,
gstate: PDFGraphicState,
stroke: bool,
fill: bool,
evenodd: bool,
path: Sequence[PathSegment],
) -> None:
"""Paint paths described in section 4.4 of the PDF reference manual"""
shape = "".join(x[0] for x in path)
if shape[:1] != "m":
# Per PDF Reference Section 4.4.1, "path construction operators may
# be invoked in any sequence, but the first one invoked must be m
# or re to begin a new subpath." Since pdf2zh.six already
# converts all `re` (rectangle) operators to their equivelent
# `mlllh` representation, paths ingested by `.paint_path(...)` that
# do not begin with the `m` operator are invalid.
pass
elif shape.count("m") > 1:
# recurse if there are multiple m's in this shape
for m in re.finditer(r"m[^m]+", shape):
subpath = path[m.start(0) : m.end(0)]
self.paint_path(gstate, stroke, fill, evenodd, subpath)
else:
# Although the 'h' command does not not literally provide a
# point-position, its position is (by definition) equal to the
# subpath's starting point.
#
# And, per Section 4.4's Table 4.9, all other path commands place
# their point-position in their final two arguments. (Any preceding
# arguments represent control points on Bézier curves.)
raw_pts = [
cast(Point, p[-2:] if p[0] != "h" else path[0][-2:]) for p in path
]
pts = [apply_matrix_pt(self.ctm, pt) for pt in raw_pts]
operators = [str(operation[0]) for operation in path]
transformed_points = [
[
apply_matrix_pt(self.ctm, (float(operand1), float(operand2)))
for operand1, operand2 in zip(operation[1::2], operation[2::2])
]
for operation in path
]
transformed_path = [
cast(PathSegment, (o, *p))
for o, p in zip(operators, transformed_points)
]
if shape in {"mlh", "ml"}:
# single line segment
#
# Note: 'ml', in conditional above, is a frequent anomaly
# that we want to support.
line = LTLine(
gstate.linewidth * matrix_scale(self.ctm),
pts[0],
pts[1],
stroke,
fill,
evenodd,
gstate.scolor,
gstate.ncolor,
original_path=transformed_path,
dashing_style=gstate.dash,
)
self.cur_item.add(line)
elif shape in {"mlllh", "mllll"}:
(x0, y0), (x1, y1), (x2, y2), (x3, y3), _ = pts
is_closed_loop = pts[0] == pts[4]
has_square_coordinates = (
x0 == x1 and y1 == y2 and x2 == x3 and y3 == y0
) or (y0 == y1 and x1 == x2 and y2 == y3 and x3 == x0)
if is_closed_loop and has_square_coordinates:
rect = LTRect(
gstate.linewidth * matrix_scale(self.ctm),
(*pts[0], *pts[2]),
stroke,
fill,
evenodd,
gstate.scolor,
gstate.ncolor,
transformed_path,
gstate.dash,
)
self.cur_item.add(rect)
else:
curve = LTCurve(
gstate.linewidth * matrix_scale(self.ctm),
pts,
stroke,
fill,
evenodd,
gstate.scolor,
gstate.ncolor,
transformed_path,
gstate.dash,
)
self.cur_item.add(curve)
else:
curve = LTCurve(
gstate.linewidth * matrix_scale(self.ctm),
pts,
stroke,
fill,
evenodd,
gstate.scolor,
gstate.ncolor,
transformed_path,
gstate.dash,
)
self.cur_item.add(curve)
def render_char(
self,
matrix: Matrix,
font: PDFFont,
fontsize: float,
scaling: float,
rise: float,
cid: int,
ncs: PDFColorSpace,
graphicstate: PDFGraphicState,
) -> float:
try:
text = font.to_unichr(cid)
assert isinstance(text, str), str(type(text))
except PDFUnicodeNotDefined:
text = self.handle_undefined_char(font, cid)
textwidth = font.char_width(cid)
textdisp = font.char_disp(cid)
item = LTChar(
matrix,
font,
fontsize,
scaling,
rise,
text,
textwidth,
textdisp,
ncs,
graphicstate,
)
self.cur_item.add(item)
item.cid = cid # hack 插入原字符编码
return item.adv
def handle_undefined_char(self, font: PDFFont, cid: int) -> str:
# log.debug("undefined: %r, %r", font, cid)
return "(cid:%d)" % cid
def receive_layout(self, ltpage: LTPage) -> None:
pass
class PDFPageAggregator(PDFLayoutAnalyzer):
def __init__(
self,
rsrcmgr: PDFResourceManager,
pageno: int = 1,
laparams: Optional[LAParams] = None,
) -> None:
PDFLayoutAnalyzer.__init__(self, rsrcmgr, pageno=pageno, laparams=laparams)
self.result: Optional[LTPage] = None
def receive_layout(self, ltpage: LTPage) -> None:
self.result = ltpage
def get_result(self) -> LTPage:
assert self.result is not None
return self.result
# Some PDFConverter children support only binary I/O
IOType = TypeVar("IOType", TextIO, BinaryIO, AnyIO)
class PDFConverter(PDFLayoutAnalyzer, Generic[IOType]):
def __init__(
self,
rsrcmgr: PDFResourceManager,
outfp: IOType,
codec: str = "utf-8",
pageno: int = 1,
laparams: Optional[LAParams] = None,
) -> None:
PDFLayoutAnalyzer.__init__(self, rsrcmgr, pageno=pageno, laparams=laparams)
self.outfp: IOType = outfp
self.codec = codec
self.outfp_binary = self._is_binary_stream(self.outfp)
@staticmethod
def _is_binary_stream(outfp: AnyIO) -> bool:
"""Test if an stream is binary or not"""
if "b" in getattr(outfp, "mode", ""):
return True
elif hasattr(outfp, "mode"):
# output stream has a mode, but it does not contain 'b'
return False
elif isinstance(outfp, io.BytesIO):
return True
elif isinstance(outfp, io.StringIO) or isinstance(outfp, io.TextIOBase):
return False
return True
class TextConverter(PDFConverter[AnyIO]):
def __init__(
self,
rsrcmgr: PDFResourceManager,
outfp: AnyIO,
codec: str = "utf-8",
pageno: int = 1,
laparams: Optional[LAParams] = None,
showpageno: bool = False,
imagewriter: Optional[ImageWriter] = None,
vfont: str = None,
vchar: str = None,
thread: int = 0,
layout={},
lang_in: str = "",
lang_out: str = "",
service: str = "",
) -> None:
super().__init__(rsrcmgr, outfp, codec=codec, pageno=pageno, laparams=laparams)
self.showpageno = showpageno
self.imagewriter = imagewriter
self.vfont = vfont
self.vchar = vchar
self.thread = thread
self.layout = layout
param = service.split(":", 1)
if param[0] == "google":
self.translator: BaseTranslator = GoogleTranslator(
service, lang_out, lang_in, None
)
elif param[0] == "deepl":
self.translator: BaseTranslator = DeepLTranslator(
service, lang_out, lang_in, None
)
elif param[0] == "deeplx":
self.translator: BaseTranslator = DeepLXTranslator(
service, lang_out, lang_in, None
)
elif param[0] == "ollama":
self.translator: BaseTranslator = OllamaTranslator(
service, lang_out, lang_in, param[1]
)
elif param[0] == "openai":
self.translator: BaseTranslator = OpenAITranslator(
service, lang_out, lang_in, param[1]
)
elif param[0] == "azure":
self.translator: BaseTranslator = AzureTranslator(
service, lang_out, lang_in, None
)
elif param[0] == "tencent":
self.translator: BaseTranslator = TencentTranslator(
service, lang_out, lang_in, None
)
else:
raise ValueError("Unsupported translation service")
def write_text(self, text: str) -> None:
text = utils.compatible_encode_method(text, self.codec, "ignore")
if self.outfp_binary:
cast(BinaryIO, self.outfp).write(text.encode())
else:
cast(TextIO, self.outfp).write(text)
# fmt: off
def receive_layout(self, ltpage: LTPage):
xt = None # 上一个字符
sstk = [] # 段落文字栈
vstk = [] # 公式符号组
vlstk = [] # 公式线条组
vfix = 0 # 公式纵向偏移
vbkt = 0 # 段落公式括号计数
pstk = [] # 段落属性栈
lstk = [] # 全局线条栈
var = [] # 公式符号组栈
varl = [] # 公式线条组栈
varf = [] # 公式纵向偏移栈
vlen = [] # 公式宽度栈
xt_cls = -1 # 上一个字符所属段落
vmax = ltpage.width / 4 # 行内公式最大宽度
ops = "" # 渲染结果
def vflag(font, char): # 匹配公式(和角标)字体
if re.match(r"\(cid:", char):
return True
# 基于字体名规则的判定
if self.vfont:
if re.match(self.vfont, font):
return True
else:
if re.match( # latex 字体
r"(CM[^R]|MS|XY|MT|BL|RM|EU|LA|RS|LINE|TeX-|rsfs|txsy|wasy|.*Mono|.*Code|.*Ital|.*Sym)",
font,
):
return True
# 基于字符集规则的判定
if self.vchar:
if re.match(self.vchar, char):
return True
else:
if (
char
and char != " " # 非空格
and (
unicodedata.category(char[0])
in ["Lm", "Mn", "Sk", "Sm", "Zl", "Zp", "Zs"] # 文字修饰符、数学符号、分隔符号
or ord(char[0]) in range(0x370, 0x400) # 希腊字母
)
):
return True
return False
############################################################
# A. 原文档解析
ptr = 0
item = list(ltpage)
while ptr < len(item):
child = item[ptr]
if isinstance(child, LTChar):
cur_v = False
fontname = child.fontname.split("+")[-1]
layout = self.layout[ltpage.pageid]
# ltpage.height 可能是 fig 里面的高度,这里统一用 layout.shape
h, w = layout.shape
# 读取当前字符在 layout 中的类别
cx, cy = np.clip(int(child.x0), 0, w - 1), np.clip(int(child.y0), 0, h - 1)
cls = layout[cy, cx]
if ( # 判定当前字符是否属于公式
cls == 0 # 1. 类别为保留区域
or (cls == xt_cls and len(sstk[-1].strip()) > 1 and child.size < pstk[-1][4] * 0.79) # 2. 角标字体,有 0.76 的角标和 0.799 的大写,这里用 0.79 取中,同时考虑首字母放大的情况
or vflag(fontname, child.get_text()) # 3. 公式字体
or (child.matrix[0] == 0 and child.matrix[3] == 0) # 4. 垂直字体
):
cur_v = True
# 判定括号组是否属于公式
if not cur_v:
if vstk and child.get_text() == "(":
cur_v = True
vbkt += 1
if vbkt and child.get_text() == ")":
cur_v = True
vbkt -= 1
if ( # 判定当前公式是否结束
not cur_v # 1. 当前字符不属于公式
or cls != xt_cls # 2. 当前字符与前一个字符不属于同一段落
or (abs(child.x0 - xt.x0) > vmax and cls != 0) # 3. 段落内换行,可能是一长串斜体的段落,也可能是段内分式换行,这里设个阈值进行区分
):
if vstk:
if ( # 根据公式右侧的文字修正公式的纵向偏移
not cur_v # 1. 当前字符不属于公式
and cls == xt_cls # 2. 当前字符与前一个字符属于同一段落
and child.x0 > max([vch.x0 for vch in vstk]) # 3. 当前字符在公式右侧
):
vfix = vstk[0].y0 - child.y0
sstk[-1] += f"$v{len(var)}$"
var.append(vstk)
varl.append(vlstk)
varf.append(vfix)
vstk = []
vlstk = []
vfix = 0
# 当前字符不属于公式或当前字符是公式的第一个字符
if not vstk:
if cls == xt_cls: # 当前字符与前一个字符属于同一段落
if child.x0 > xt.x1 + 1: # 添加行内空格
sstk[-1] += " "
elif child.x1 < xt.x0: # 添加换行空格并标记原文段落存在换行
sstk[-1] += " "
pstk[-1][6] = True
else: # 根据当前字符构建一个新的段落
sstk.append("")
pstk.append([child.y0, child.x0, child.x0, child.x0, child.size, child.font, False])
if not cur_v: # 文字入栈
if ( # 根据当前字符修正段落属性
child.size > pstk[-1][4] / 0.79 # 1. 当前字符显著比段落字体大
or len(sstk[-1].strip()) == 1 # 2. 当前字符为段落第二个文字(考虑首字母放大的情况)
or vflag(pstk[-1][5].fontname.split("+")[-1], "") # 3. 段落字体为公式字体
or re.match( # 4. 段落字体为粗体
r"(.*Medi|.*Bold)",
pstk[-1][5].fontname.split("+")[-1],
re.IGNORECASE,
)
):
pstk[-1][0] -= child.size - pstk[-1][4] # hack 这个段落纵向位置的修正有问题,不过先凑合用吧
pstk[-1][4] = child.size
pstk[-1][5] = child.font
sstk[-1] += child.get_text()
else: # 公式入栈
if ( # 根据公式左侧的文字修正公式的纵向偏移
not vstk # 1. 当前字符是公式的第一个字符
and cls == xt_cls # 2. 当前字符与前一个字符属于同一段落
and child.x0 > xt.x0 # 3. 前一个字符在公式左侧
):
vfix = child.y0 - xt.y0
vstk.append(child)
# 更新段落边界,因为段落内换行之后可能是公式开头,所以要在外边处理
pstk[-1][2] = min(pstk[-1][2], child.x0)
pstk[-1][3] = max(pstk[-1][3], child.x1)
# 更新上一个字符
xt = child
xt_cls = cls
elif isinstance(child, LTFigure): # 图表
pass
elif isinstance(child, LTLine): # 线条
layout = self.layout[ltpage.pageid]
# ltpage.height 可能是 fig 里面的高度,这里统一用 layout.shape
h, w = layout.shape
# 读取当前线条在 layout 中的类别
cx, cy = np.clip(int(child.x0), 0, w - 1), np.clip(int(child.y0), 0, h - 1)
cls = layout[cy, cx]
if vstk and cls == xt_cls: # 公式线条
vlstk.append(child)
else: # 全局线条
lstk.append(child)
else:
pass
ptr += 1
# 处理结尾
if vstk: # 公式出栈
sstk[-1] += f"$v{len(var)}$"
var.append(vstk)
varl.append(vlstk)
varf.append(vfix)
log.debug("\n==========[VSTACK]==========\n")
for id, v in enumerate(var): # 计算公式宽度
l = max([vch.x1 for vch in v]) - v[0].x0
log.debug(f'< {l:.1f} {v[0].x0:.1f} {v[0].y0:.1f} {v[0].cid} {v[0].fontname} {len(varl[id])} > $v{id}$ = {"".join([ch.get_text() for ch in v])}')
vlen.append(l)
############################################################
# B. 段落翻译
log.debug("\n==========[SSTACK]==========\n")
hash_key = cache.deterministic_hash("PDFMathTranslate")
cache.create_cache(hash_key)
@retry(wait=wait_fixed(1))
def worker(s): # 多线程翻译
try:
hash_key_paragraph = cache.deterministic_hash(
(s, str(self.translator))
)
new = cache.load_paragraph(hash_key, hash_key_paragraph) # 查询缓存
if new is None:
new = self.translator.translate(s)
new = remove_control_characters(new)
cache.write_paragraph(hash_key, hash_key_paragraph, new)
return new
except BaseException as e:
if log.isEnabledFor(logging.DEBUG):
log.exception(e)
else:
log.exception(e, exc_info=False)
raise e
with concurrent.futures.ThreadPoolExecutor(
max_workers=self.thread
) as executor:
news = list(executor.map(worker, sstk))
############################################################
# C. 新文档排版
def raw_string(fcur, cstk): # 编码字符串
if isinstance(self.fontmap[fcur], PDFCIDFont): # 判断编码长度
return "".join(["%04x" % ord(c) for c in cstk])
else:
return "".join(["%02x" % ord(c) for c in cstk])
_x, _y = 0, 0
for id, new in enumerate(news):
tx = x = pstk[id][1] # 段落初始横坐标
y = pstk[id][0] # 段落上边界
lt = pstk[id][2] # 段落左边界
rt = pstk[id][3] # 段落右边界
size = pstk[id][4] # 段落字体大小
font = pstk[id][5] # 段落字体
lb = pstk[id][6] # 段落属性
cstk = "" # 当前文字栈
fcur = fcur_ = None # 当前字体
ptr = 0
log.debug(f"< {y} {x} {lt} {rt} {size} {font.fontname} {lb} > {sstk[id]} | {new}")
while True:
if ptr == len(new): # 到达段落结尾
if cstk:
ops += f"/{fcur} {size:f} Tf 1 0 0 1 {tx:f} {y:f} Tm [<{raw_string(fcur, cstk)}>] TJ "
break
vy_regex = re.match(
r"\$?\s*v([\d\s]+)\$", new[ptr:], re.IGNORECASE
) # 匹配 $vn$ 公式标记,前面的 $ 有的时候会被丢掉
mod = False # 当前公式是否为文字修饰符
if vy_regex: # 加载公式
ptr += len(vy_regex.group(0))
try:
vid = int(vy_regex.group(1).replace(" ", ""))
adv = vlen[vid]
except Exception:
continue # 翻译器可能会自动补个越界的公式标记
if len(var[vid]) == 1 and unicodedata.category(var[vid][0].get_text()[0]) in ["Lm", "Mn", "Sk"]: # 文字修饰符
mod = True
else: # 加载文字
ch = new[ptr]
# if font.char_width(ord(ch)):
fcur_ = None
# 原字体编码容易出问题,这里直接放弃掉
# try:
# if font.widths.get(ord(ch)) and font.to_unichr(ord(ch))==ch:
# fcur_=self.fontid[font] # 原字体
# except:
# pass
try:
if fcur_ is None and self.fontmap["tiro"].to_unichr(ord(ch)) == ch:
fcur_ = "tiro" # 默认英文字体
except Exception:
pass
if fcur_ is None:
fcur_ = "china-ss" # 默认中文字体
# print(self.fontid[font],fcur_,ch,font.char_width(ord(ch)))
adv = self.fontmap[fcur_].char_width(ord(ch)) * size
ptr += 1
if ( # 输出文字缓冲区
fcur_ != fcur # 1. 字体更新
or vy_regex # 2. 插入公式
or x + adv > rt + 0.1 * size # 3. 到达右边界(可能一整行都被符号化,这里需要考虑浮点误差)
):
if cstk:
ops += f"/{fcur} {size:f} Tf 1 0 0 1 {tx:f} {y:f} Tm [<{raw_string(fcur, cstk)}>] TJ "
cstk = ""
if lb and x + adv > rt + 0.1 * size: # 到达右边界且原文段落存在换行
x = lt
lang_space = {"zh-CN": 1.4, "zh-TW": 1.4, "ja": 1.1, "ko": 1.2, "en": 1.2} # CJK
y -= size * lang_space.get(self.translator.lang_out, 1.1) # 小语种大多适配 1.1
if vy_regex: # 插入公式
fix = 0
if fcur is not None: # 段落内公式修正纵向偏移
fix = varf[vid]
for vch in var[vid]: # 排版公式字符
vc = chr(vch.cid)
ops += f"/{self.fontid[vch.font]} {vch.size:f} Tf 1 0 0 1 {x + vch.x0 - var[vid][0].x0:f} {fix + y + vch.y0 - var[vid][0].y0:f} Tm [<{raw_string(self.fontid[vch.font], vc)}>] TJ "
if log.isEnabledFor(logging.DEBUG):
lstk.append(LTLine(0.1, (_x, _y), (x + vch.x0 - var[vid][0].x0, fix + y + vch.y0 - var[vid][0].y0)))
_x, _y = x + vch.x0 - var[vid][0].x0, fix + y + vch.y0 - var[vid][0].y0
for l in varl[vid]: # 排版公式线条
if l.linewidth < 5: # hack 有的文档会用粗线条当图片背景
ops += f"ET q 1 0 0 1 {l.pts[0][0] + x - var[vid][0].x0:f} {l.pts[0][1] + fix + y - var[vid][0].y0:f} cm [] 0 d 0 J {l.linewidth:f} w 0 0 m {l.pts[1][0] - l.pts[0][0]:f} {l.pts[1][1] - l.pts[0][1]:f} l S Q BT "
else: # 插入文字缓冲区
if not cstk: # 单行开头
tx = x
if x == lt and ch == " ": # 消除段落换行空格
adv = 0
else:
cstk += ch
else:
cstk += ch
if mod: # 文字修饰符
adv = 0
fcur = fcur_
x += adv
if log.isEnabledFor(logging.DEBUG):
lstk.append(LTLine(0.1, (_x, _y), (x, y)))
_x, _y = x, y
for l in lstk: # 排版全局线条
if l.linewidth < 5: # hack 有的文档会用粗线条当图片背景
ops += f"ET q 1 0 0 1 {l.pts[0][0]:f} {l.pts[0][1]:f} cm [] 0 d 0 J {l.linewidth:f} w 0 0 m {l.pts[1][0] - l.pts[0][0]:f} {l.pts[1][1] - l.pts[0][1]:f} l S Q BT "
ops = f"BT {ops}ET "
return ops
# Some dummy functions to save memory/CPU when all that is wanted
# is text. This stops all the image and drawing output from being
# recorded and taking up RAM.
def render_image(self, name: str, stream: PDFStream) -> None:
if self.imagewriter is not None:
PDFConverter.render_image(self, name, stream)
# def paint_path(
# self,
# gstate: PDFGraphicState,
# stroke: bool,
# fill: bool,
# evenodd: bool,
# path: Sequence[PathSegment],
# ) -> None:
# pass
class HTMLConverter(PDFConverter[AnyIO]):
RECT_COLORS = {
"figure": "yellow",
"textline": "magenta",
"textbox": "cyan",
"textgroup": "red",
"curve": "black",
"page": "gray",
}
TEXT_COLORS = {
"textbox": "blue",
"char": "black",
}
def __init__(
self,
rsrcmgr: PDFResourceManager,
outfp: AnyIO,
codec: str = "utf-8",
pageno: int = 1,
laparams: Optional[LAParams] = None,
scale: float = 1,
fontscale: float = 1.0,
layoutmode: str = "normal",
showpageno: bool = True,
pagemargin: int = 50,
imagewriter: Optional[ImageWriter] = None,
debug: int = 0,
rect_colors: Optional[Dict[str, str]] = None,
text_colors: Optional[Dict[str, str]] = None,
) -> None:
PDFConverter.__init__(
self,
rsrcmgr,
outfp,
codec=codec,
pageno=pageno,
laparams=laparams,
)
# write() assumes a codec for binary I/O, or no codec for text I/O.
if self.outfp_binary and not self.codec:
raise PDFValueError("Codec is required for a binary I/O output")
if not self.outfp_binary and self.codec:
raise PDFValueError("Codec must not be specified for a text I/O output")
if text_colors is None:
text_colors = {"char": "black"}
if rect_colors is None:
rect_colors = {"curve": "black", "page": "gray"}
self.scale = scale
self.fontscale = fontscale
self.layoutmode = layoutmode
self.showpageno = showpageno
self.pagemargin = pagemargin
self.imagewriter = imagewriter
self.rect_colors = rect_colors
self.text_colors = text_colors
if debug:
self.rect_colors.update(self.RECT_COLORS)
self.text_colors.update(self.TEXT_COLORS)
self._yoffset: float = self.pagemargin
self._font: Optional[Tuple[str, float]] = None
self._fontstack: List[Optional[Tuple[str, float]]] = []
self.write_header()
def write(self, text: str) -> None:
if self.codec:
cast(BinaryIO, self.outfp).write(text.encode(self.codec))
else:
cast(TextIO, self.outfp).write(text)
def write_header(self) -> None:
self.write("<html><head>\n")
if self.codec:
s = (
'<meta http-equiv="Content-Type" content="text/html; '
'charset=%s">\n' % self.codec
)
else:
s = '<meta http-equiv="Content-Type" content="text/html">\n'
self.write(s)
self.write("</head><body>\n")
def write_footer(self) -> None:
page_links = [f'<a href="#{i}">{i}</a>' for i in range(1, self.pageno)]
s = '<div style="position:absolute; top:0px;">Page: %s</div>\n' % ", ".join(
page_links,
)
self.write(s)
self.write("</body></html>\n")
def write_text(self, text: str) -> None:
self.write(enc(text))
def place_rect(
self,
color: str,
borderwidth: int,
x: float,
y: float,
w: float,
h: float,
) -> None:
color2 = self.rect_colors.get(color)
if color2 is not None:
s = (
'<span style="position:absolute; border: %s %dpx solid; '
'left:%dpx; top:%dpx; width:%dpx; height:%dpx;"></span>\n'
% (
color2,
borderwidth,
x * self.scale,
(self._yoffset - y) * self.scale,
w * self.scale,
h * self.scale,
)
)
self.write(s)
def place_border(self, color: str, borderwidth: int, item: LTComponent) -> None:
self.place_rect(color, borderwidth, item.x0, item.y1, item.width, item.height)
def place_image(
self,
item: LTImage,
borderwidth: int,
x: float,
y: float,
w: float,
h: float,
) -> None:
if self.imagewriter is not None:
name = self.imagewriter.export_image(item)
s = (
'<img src="%s" border="%d" style="position:absolute; '
'left:%dpx; top:%dpx;" width="%d" height="%d" />\n'
% (
enc(name),
borderwidth,
x * self.scale,
(self._yoffset - y) * self.scale,
w * self.scale,
h * self.scale,
)
)
self.write(s)
def place_text(
self,
color: str,
text: str,
x: float,
y: float,
size: float,
) -> None:
color2 = self.text_colors.get(color)
if color2 is not None:
s = (
'<span style="position:absolute; color:%s; left:%dpx; '
'top:%dpx; font-size:%dpx;">'
% (
color2,
x * self.scale,
(self._yoffset - y) * self.scale,
size * self.scale * self.fontscale,
)
)
self.write(s)
self.write_text(text)
self.write("</span>\n")
def begin_div(
self,
color: str,
borderwidth: int,
x: float,
y: float,
w: float,
h: float,
writing_mode: str = "False",
) -> None:
self._fontstack.append(self._font)
self._font = None
s = (
'<div style="position:absolute; border: %s %dpx solid; '
"writing-mode:%s; left:%dpx; top:%dpx; width:%dpx; "
'height:%dpx;">'
% (
color,
borderwidth,
writing_mode,
x * self.scale,
(self._yoffset - y) * self.scale,
w * self.scale,
h * self.scale,
)
)
self.write(s)
def end_div(self, color: str) -> None:
if self._font is not None:
self.write("</span>")
self._font = self._fontstack.pop()
self.write("</div>")
def put_text(self, text: str, fontname: str, fontsize: float) -> None:
font = (fontname, fontsize)
if font != self._font:
if self._font is not None:
self.write("</span>")
# Remove subset tag from fontname, see PDF Reference 5.5.3
fontname_without_subset_tag = fontname.split("+")[-1]
self.write(
'<span style="font-family: %s; font-size:%dpx">'
% (fontname_without_subset_tag, fontsize * self.scale * self.fontscale),
)
self._font = font
self.write_text(text)
def put_newline(self) -> None:
self.write("<br>")
def receive_layout(self, ltpage: LTPage) -> None:
def show_group(item: Union[LTTextGroup, TextGroupElement]) -> None:
if isinstance(item, LTTextGroup):
self.place_border("textgroup", 1, item)
for child in item:
show_group(child)
def render(item: LTItem) -> None:
child: LTItem
if isinstance(item, LTPage):
self._yoffset += item.y1
self.place_border("page", 1, item)
if self.showpageno:
self.write(
'<div style="position:absolute; top:%dpx;">'
% ((self._yoffset - item.y1) * self.scale),
)
self.write(
f'<a name="{item.pageid}">Page {item.pageid}</a></div>\n',
)
for child in item:
render(child)
if item.groups is not None:
for group in item.groups:
show_group(group)
elif isinstance(item, LTCurve):
self.place_border("curve", 1, item)
elif isinstance(item, LTFigure):
self.begin_div("figure", 1, item.x0, item.y1, item.width, item.height)
for child in item:
render(child)
self.end_div("figure")
elif isinstance(item, LTImage):
self.place_image(item, 1, item.x0, item.y1, item.width, item.height)
elif self.layoutmode == "exact":
if isinstance(item, LTTextLine):
self.place_border("textline", 1, item)
for child in item:
render(child)
elif isinstance(item, LTTextBox):
self.place_border("textbox", 1, item)
self.place_text(
"textbox",
str(item.index + 1),
item.x0,
item.y1,
20,
)
for child in item:
render(child)
elif isinstance(item, LTChar):
self.place_border("char", 1, item)
self.place_text(
"char",
item.get_text(),
item.x0,
item.y1,
item.size,
)
elif isinstance(item, LTTextLine):
for child in item:
render(child)
if self.layoutmode != "loose":
self.put_newline()
elif isinstance(item, LTTextBox):
self.begin_div(
"textbox",
1,
item.x0,
item.y1,
item.width,
item.height,
item.get_writing_mode(),
)
for child in item:
render(child)
self.end_div("textbox")
elif isinstance(item, LTChar):
fontname = make_compat_str(item.fontname)
self.put_text(item.get_text(), fontname, item.size)
elif isinstance(item, LTText):
self.write_text(item.get_text())
render(ltpage)
self._yoffset += self.pagemargin
def close(self) -> None:
self.write_footer()
class XMLConverter(PDFConverter[AnyIO]):
CONTROL = re.compile("[\x00-\x08\x0b-\x0c\x0e-\x1f]")
def __init__(
self,
rsrcmgr: PDFResourceManager,
outfp: AnyIO,
codec: str = "utf-8",
pageno: int = 1,
laparams: Optional[LAParams] = None,
imagewriter: Optional[ImageWriter] = None,
stripcontrol: bool = False,
) -> None:
PDFConverter.__init__(
self,
rsrcmgr,
outfp,
codec=codec,
pageno=pageno,
laparams=laparams,
)
# write() assumes a codec for binary I/O, or no codec for text I/O.
if self.outfp_binary == (not self.codec):
raise PDFValueError("Codec is required for a binary I/O output")
self.imagewriter = imagewriter
self.stripcontrol = stripcontrol
self.write_header()
def write(self, text: str) -> None:
if self.codec:
cast(BinaryIO, self.outfp).write(text.encode(self.codec))
else:
cast(TextIO, self.outfp).write(text)
def write_header(self) -> None:
if self.codec:
self.write('<?xml version="1.0" encoding="%s" ?>\n' % self.codec)
else:
self.write('<?xml version="1.0" ?>\n')
self.write("<pages>\n")
def write_footer(self) -> None:
self.write("</pages>\n")
def write_text(self, text: str) -> None:
if self.stripcontrol:
text = self.CONTROL.sub("", text)
self.write(enc(text))
def receive_layout(self, ltpage: LTPage) -> None:
def show_group(item: LTItem) -> None:
if isinstance(item, LTTextBox):
self.write(
'<textbox id="%d" bbox="%s" />\n'
% (item.index, bbox2str(item.bbox)),
)
elif isinstance(item, LTTextGroup):
self.write('<textgroup bbox="%s">\n' % bbox2str(item.bbox))
for child in item:
show_group(child)
self.write("</textgroup>\n")
def render(item: LTItem) -> None:
child: LTItem
if isinstance(item, LTPage):
s = '<page id="%s" bbox="%s" rotate="%d">\n' % (
item.pageid,
bbox2str(item.bbox),
item.rotate,
)
self.write(s)
for child in item:
render(child)
if item.groups is not None:
self.write("<layout>\n")
for group in item.groups:
show_group(group)
self.write("</layout>\n")
self.write("</page>\n")
elif isinstance(item, LTLine):
s = '<line linewidth="%d" bbox="%s" />\n' % (
item.linewidth,
bbox2str(item.bbox),
)
self.write(s)
elif isinstance(item, LTRect):
s = '<rect linewidth="%d" bbox="%s" />\n' % (
item.linewidth,
bbox2str(item.bbox),
)
self.write(s)
elif isinstance(item, LTCurve):
s = '<curve linewidth="%d" bbox="%s" pts="%s"/>\n' % (
item.linewidth,
bbox2str(item.bbox),
item.get_pts(),
)
self.write(s)
elif isinstance(item, LTFigure):
s = f'<figure name="{item.name}" bbox="{bbox2str(item.bbox)}">\n'
self.write(s)
for child in item:
render(child)
self.write("</figure>\n")
elif isinstance(item, LTTextLine):
self.write('<textline bbox="%s">\n' % bbox2str(item.bbox))
for child in item:
render(child)
self.write("</textline>\n")
elif isinstance(item, LTTextBox):
wmode = ""
if isinstance(item, LTTextBoxVertical):
wmode = ' wmode="vertical"'
s = '<textbox id="%d" bbox="%s"%s>\n' % (
item.index,
bbox2str(item.bbox),
wmode,
)
self.write(s)
for child in item:
render(child)
self.write("</textbox>\n")
elif isinstance(item, LTChar):
s = (
'<text font="%s" bbox="%s" colourspace="%s" '
'ncolour="%s" size="%.3f">'
% (
enc(item.fontname),
bbox2str(item.bbox),
item.ncs.name,
item.graphicstate.ncolor,
item.size,
)
)
self.write(s)
self.write_text(item.get_text())
self.write("</text>\n")
elif isinstance(item, LTText):
self.write("<text>%s</text>\n" % item.get_text())
elif isinstance(item, LTImage):
if self.imagewriter is not None:
name = self.imagewriter.export_image(item)
self.write(
'<image src="%s" width="%d" height="%d" />\n'
% (enc(name), item.width, item.height),
)
else:
self.write(
'<image width="%d" height="%d" />\n'
% (item.width, item.height),
)
else:
assert False, str(("Unhandled", item))
render(ltpage)
def close(self) -> None:
self.write_footer()
class HOCRConverter(PDFConverter[AnyIO]):
"""Extract an hOCR representation from explicit text information within a PDF."""
# Where text is being extracted from a variety of types of PDF within a
# business process, those PDFs where the text is only present in image
# form will need to be analysed using an OCR tool which will typically
# output hOCR. This converter extracts the explicit text information from
# those PDFs that do have it and uses it to genxerate a basic hOCR
# representation that is designed to be used in conjunction with the image
# of the PDF in the same way as genuine OCR output would be, but without the
# inevitable OCR errors.
# The converter does not handle images, diagrams or text colors.
# In the examples processed by the contributor it was necessary to set
# LAParams.all_texts to True.
CONTROL = re.compile(r"[\x00-\x08\x0b-\x0c\x0e-\x1f]")
def __init__(
self,
rsrcmgr: PDFResourceManager,
outfp: AnyIO,
codec: str = "utf8",
pageno: int = 1,
laparams: Optional[LAParams] = None,
stripcontrol: bool = False,
):
PDFConverter.__init__(
self,
rsrcmgr,
outfp,
codec=codec,
pageno=pageno,
laparams=laparams,
)
self.stripcontrol = stripcontrol
self.within_chars = False
self.write_header()
def bbox_repr(self, bbox: Rect) -> str:
(in_x0, in_y0, in_x1, in_y1) = bbox
# PDF y-coordinates are the other way round from hOCR coordinates
out_x0 = int(in_x0)
out_y0 = int(self.page_bbox[3] - in_y1)
out_x1 = int(in_x1)
out_y1 = int(self.page_bbox[3] - in_y0)
return f"bbox {out_x0} {out_y0} {out_x1} {out_y1}"
def write(self, text: str) -> None:
if self.codec:
encoded_text = text.encode(self.codec)
cast(BinaryIO, self.outfp).write(encoded_text)
else:
cast(TextIO, self.outfp).write(text)
def write_header(self) -> None:
if self.codec:
self.write(
"<html xmlns='http://www.w3.org/1999/xhtml' "
"xml:lang='en' lang='en' charset='%s'>\n" % self.codec,
)
else:
self.write(
"<html xmlns='http://www.w3.org/1999/xhtml' "
"xml:lang='en' lang='en'>\n",
)
self.write("<head>\n")
self.write("<title></title>\n")
self.write(
"<meta http-equiv='Content-Type' content='text/html;charset=utf-8' />\n",
)
self.write(
"<meta name='ocr-system' content='pdf2zh.six HOCR Converter' />\n",
)
self.write(
" <meta name='ocr-capabilities'"
" content='ocr_page ocr_block ocr_line ocrx_word'/>\n",
)
self.write("</head>\n")
self.write("<body>\n")
def write_footer(self) -> None:
self.write("<!-- comment in the following line to debug -->\n")
self.write(
"<!--script src='https://unpkg.com/hocrjs'></script--></body></html>\n",
)
def write_text(self, text: str) -> None:
if self.stripcontrol:
text = self.CONTROL.sub("", text)
self.write(text)
def write_word(self) -> None:
if len(self.working_text) > 0:
bold_and_italic_styles = ""
if "Italic" in self.working_font:
bold_and_italic_styles = "font-style: italic; "
if "Bold" in self.working_font:
bold_and_italic_styles += "font-weight: bold; "
self.write(
"<span style='font:\"%s\"; font-size:%d; %s' "
"class='ocrx_word' title='%s; x_font %s; "
"x_fsize %d'>%s</span>"
% (
(
self.working_font,
self.working_size,
bold_and_italic_styles,
self.bbox_repr(self.working_bbox),
self.working_font,
self.working_size,
self.working_text.strip(),
)
),
)
self.within_chars = False
def receive_layout(self, ltpage: LTPage) -> None:
def render(item: LTItem) -> None:
if self.within_chars and isinstance(item, LTAnno):
self.write_word()
if isinstance(item, LTPage):
self.page_bbox = item.bbox
self.write(
"<div class='ocr_page' id='%s' title='%s'>\n"
% (item.pageid, self.bbox_repr(item.bbox)),
)
for child in item:
render(child)
self.write("</div>\n")
elif isinstance(item, LTTextLine):
self.write(
"<span class='ocr_line' title='%s'>" % (self.bbox_repr(item.bbox)),
)
for child_line in item:
render(child_line)
self.write("</span>\n")
elif isinstance(item, LTTextBox):
self.write(
"<div class='ocr_block' id='%d' title='%s'>\n"
% (item.index, self.bbox_repr(item.bbox)),
)
for child in item:
render(child)
self.write("</div>\n")
elif isinstance(item, LTChar):
if not self.within_chars:
self.within_chars = True
self.working_text = item.get_text()
self.working_bbox = item.bbox
self.working_font = item.fontname
self.working_size = item.size
elif len(item.get_text().strip()) == 0:
self.write_word()
self.write(item.get_text())
else:
if (
self.working_bbox[1] != item.bbox[1]
or self.working_font != item.fontname
or self.working_size != item.size
):
self.write_word()
self.working_bbox = item.bbox
self.working_font = item.fontname
self.working_size = item.size
self.working_text += item.get_text()
self.working_bbox = (
self.working_bbox[0],
self.working_bbox[1],
item.bbox[2],
self.working_bbox[3],
)
render(ltpage)
def close(self) -> None:
self.write_footer()