add paragraph parser

Author: meldonization

depdf/__init__.py

@@ -1,9 +1,15 @@
-from depdf.api import convert_pdf_to_html, convert_pdf_to_html_by_page
+from depdf.api import *
+from depdf.config import Config
 from depdf.pdf import DePDF
+from depdf.page import DePage
 from depdf.version import __version__
 
 __all__ = [
-    'convert_pdf_to_html',
-    'convert_pdf_to_html_by_page',
+    'Config',
     'DePDF',
+    'DePage',
+    'convert_pdf_to_html',
+    'convert_page_to_html',
+    'extract_page_tables',
+    'extract_page_paragraphs',
 ]

depdf/api.py

@@ -5,6 +5,7 @@
 from depdf.error import PDFTypeError
 from depdf.log import logger_init
 from depdf.pdf import DePDF
+from depdf.page import DePage
 
 log = logger_init(__name__)
 
@@ -30,47 +31,46 @@ def wrapper(pdf_file_path, *args, **kwargs):
 
 
 @api_load_pdf
-def convert_pdf_to_html(pdf_file_path, **kwargs):
+def convert_pdf_to_html(pdf_file, **kwargs):
     """
-    :param pdf_file_path: pdf file absolute path
+    :param pdf_file: pdf file absolute path
     :param kwargs: config keyword arguments
     :return:
     """
-    html = []
-    return html
+    return pdf_file.html
 
 
 @api_load_pdf
-def convert_pdf_to_html_by_page(pdf_file_path, pid, **kwargs):
+def convert_page_to_html(pdf_file, pid, **kwargs):
     """
-    :param pdf_file_path: pdf file absolute path
+    :param pdf_file: pdf file absolute path
     :param pid: page number start from 1
     :param kwargs: config keyword arguments
     :return:
     """
-    html_page = ''
-    return html_page
+    page = pdf_file.pages[pid - 1]
+    return page.html
 
 
 @api_load_pdf
-def extract_page_tables(pdf_file_path, pid, **kwargs):
+def extract_page_tables(pdf_file, pid, **kwargs):
     """
-    :param pdf_file_path: pdf file absolute path
+    :param pdf_file: pdf file absolute path
     :param pid: page number start from 1
     :param kwargs: config keyword arguments
     :return:
     """
-    tables = []
-    return tables
+    page = pdf_file.pages[pid - 1]
+    return page.tables
 
 
 @api_load_pdf
-def extract_page_paragraphs(pdf_file_path, pid, **kwargs):
+def extract_page_paragraphs(pdf_file, pid, **kwargs):
     """
-    :param pdf_file_path: pdf file absolute path
+    :param pdf_file: pdf file absolute path
     :param pid: page number start from 1
     :param kwargs: config keyword arguments
     :return:
     """
-    paragraphs = []
-    return paragraphs
+    page = pdf_file.pages[pid - 1]
+    return page.paragraphs

depdf/base.py

@@ -11,6 +11,9 @@ class Box(object):
     bottom = Decimal(0)
     _bbox = (x0, top, x1, bottom)
 
+    def __repr__(self):
+        return '<depdf.Box: {}>'.format(tuple(self.bbox))
+
     @property
     def width(self):
         return self.x1 - self.x0
@@ -27,7 +30,7 @@ def bbox(self):
     def bbox(self, value):
         if value is not None:
             bbox = self.normalize_bbox(value)
-            (self.x0, self.top, self.x1, self.bottom) = bbox
+            self.x0, self.top, self.x1, self.bottom = bbox
             self._bbox = bbox
 
     @staticmethod
@@ -36,7 +39,9 @@ def normalize_bbox(bbox):
             raise BoxValueError(bbox)
         if isinstance(bbox, str):
             raise BoxValueError(bbox)
-        bbox = (Decimal(i) for i in bbox)
+        if len(bbox) != 4:
+            raise BoxValueError(bbox)
+        bbox = [Decimal(i) for i in bbox]
         return bbox
 
 
@@ -86,3 +91,10 @@ def refresh(self):
     def reset(self):
         pass
 
+
+class InnerWrapper(Base):
+    _inner_objects = []
+
+    @property
+    def inner_objects(self):
+        return [obj.to_dict if hasattr(obj, 'to_dict') else obj for obj in self._inner_objects]

depdf/components/paragraph.py

@@ -1,31 +1,36 @@
-from depdf.base import Base, Box
+from depdf.base import Box, InnerWrapper
 from depdf.config import check_config
 from depdf.log import logger_init
+from depdf.utils import calc_bbox, construct_style
 
 log = logger_init(__name__)
 
 
-class Paragraph(Base, Box):
+class Paragraph(InnerWrapper, Box):
     object_type = 'paragraph'
 
     @check_config
-    def __init__(self, bbox=None, text='', pid=1, para_idx=1, config=None, inner_object=None):
-        self.bbox = bbox
+    def __init__(self, bbox=None, text='', pid=1, para_idx=1, config=None, inner_objects=None, style=None):
         para_id = 'page-{pid}-paragraph-{para_id}'.format(pid=pid, para_id=para_idx)
         para_class = '{para_class} page-{pid}'.format(para_class=getattr(config, 'paragraph_class'), pid=pid)
-        html = '<p id="{para_id}" class="{para_class}">'.format(
-            para_id=para_id, para_class=para_class
+        style = construct_style(style=style)
+        html = '<p id="{para_id}" class="{para_class}"{style}>'.format(
+            para_id=para_id, para_class=para_class, style=style
         )
+        self.pid = pid
+        self.para_id = para_idx
+        self.bbox = bbox
         if text:
             self.text = text
             html += str(text)
         else:
-            self._inner_object = [inner_object]
-            for obj in inner_object:
-                self.html += getattr(obj, 'html', '')
+            if bbox is None:
+                self.bbox = calc_bbox(inner_objects)
+            self._inner_objects = inner_objects
+            for obj in inner_objects:
+                html += getattr(obj, 'html', '')
         html += '</p>'
         self.html = html
 
-    @property
-    def inner_object(self):
-        return [obj.to_dict if hasattr(obj, 'to_dict') else obj for obj in self._inner_object]
+    def __repr__(self):
+        return '<depdf.Paragraph: ({}, {})>'.format(self.pid, self.para_id)

depdf/components/span.py

@@ -1,6 +1,7 @@
 from depdf.base import Base, Box
 from depdf.config import check_config
 from depdf.log import logger_init
+from depdf.utils import construct_style
 
 log = logger_init(__name__)
 
@@ -9,10 +10,11 @@ class Span(Base, Box):
     object_type = 'span'
 
     @check_config
-    def __init__(self, bbox=None, span_text='', pid=1, config=None):
+    def __init__(self, bbox=None, span_text='', config=None, style=None):
         self.bbox = bbox
         self.text = span_text
         span_class = getattr(config, 'span_class')
-        self.html = '<span class="{span_class} page-{pid}">{span_text}</span>'.format(
-            span_class=span_class, pid=pid, span_text=span_text
+        style = construct_style(style=style)
+        self.html = '<span class="{span_class}{style}">{span_text}</span>'.format(
+            span_class=span_class, span_text=span_text, style=style
         )

depdf/components/table.py

@@ -1,28 +1,24 @@
-from depdf.base import Base, Box
+from depdf.base import Base, Box, InnerWrapper
 from depdf.config import check_config
 from depdf.log import logger_init
+from depdf.utils import calc_bbox
 
 log = logger_init(__name__)
 
 
-class Cell(Base, Box):
+class Cell(InnerWrapper, Box):
     object_type = 'cell'
 
-    def __init__(self, bbox=None, text='', font_size=14, inner_object=None):
+    def __init__(self, bbox=None, text='', inner_objects=None):
         self.bbox = bbox
-        self.fs = font_size
         if text:
             self.text = text
             self.html = text
         else:
-            self._inner_object = inner_object
-            for obj in inner_object:
+            self._inner_objects = inner_objects
+            for obj in inner_objects:
                 self.html += getattr(obj, 'html', '')
 
-    @property
-    def inner_object(self):
-        return self._inner_object.to_dict if hasattr(self._inner_object, 'to_dict') else self._inner_object
-
 
 class Table(Base, Box):
     object_type = 'table'
@@ -33,24 +29,10 @@ def __init__(self, rows, pid=1, tid=1, config=None, bbox=None):
         self.tid = tid
         self.rows = rows
         self.config = config
-        self.bbox = bbox if bbox else self.calc_table_bbox_by_rows(rows)
+        self.bbox = bbox if bbox else calc_bbox(rows)
 
-    @staticmethod
-    def calc_table_bbox_by_rows(rows):
-        x0_list, top_list, x1_list, bottom_list = [], [], [], []
-        for row in rows:
-            for cell in row:
-                x0_list.append(cell.x0)
-                top_list.append(cell.top)
-                x1_list.append(cell.x1)
-                bottom_list.append(cell.bottom)
-        bbox = (
-            min(x0_list),
-            min(top_list),
-            max(x1_list),
-            max(bottom_list),
-        )
-        return bbox
+    def __repr__(self):
+        return '<depdf.Table: ({}, {})>'.format(self.pid, self.tid)
 
     @property
     def to_dict(self):
@@ -63,16 +45,21 @@ def to_dict(self):
         ]
         return table_dict
 
+    @property
+    def html(self):
+        if not self._html and hasattr(self, 'to_html'):
+            return self.to_html
+        return self._html
+
     @property
     def to_html(self):
         table_class = getattr(self.config, 'table_class')
         table_cell_merge_tolerance = getattr(self.config, 'table_cell_merge_tolerance')
         skip_empty_table = getattr(self.config, 'skip_empty_table')
-        self.html = convert_table_to_html(
+        return convert_table_to_html(
             self.to_dict, pid=self.pid, tid=self.tid, tc_mt=table_cell_merge_tolerance,
             table_class=table_class, skip_et=skip_empty_table
         )
-        return self.html
 
 
 def gen_column_cell_sizes(t):

depdf/components/text.py

@@ -1,9 +1,10 @@
-from depdf.base import Base
+from depdf.base import Base, Box
 
 
-class Text(Base):
+class Text(Base, Box):
     object_type = 'text'
 
-    def __init__(self, text):
+    def __init__(self, bbox='', text=''):
+        self.bbox = bbox
         self.text = text
         self.html = text

depdf/config.py

@@ -61,6 +61,7 @@ class Config(Base):
     table_class = DEFAULT_TABLE_CLASS
     pdf_class = DEFAULT_PDF_CLASS
     image_class = DEFAULT_IMAGE_CLASS
+    page_class = DEFAULT_PAGE_CLASS
 
     def __init__(self, **kwargs):
         # set log level automatically if debug mode enabled
@@ -71,10 +72,14 @@ def __init__(self, **kwargs):
 
         # add configuration parameters
         self.update(**kwargs)
+        self._kwargs = kwargs
 
         # set logging level by log_level parameter
         logging.getLogger('depdf').setLevel(self.log_level)
 
+    def __repr__(self):
+        return '<depdf.Config: {}>'.format(self._kwargs)
+
     def update(self, **kwargs):
         for key, value in kwargs.items():
             if hasattr(self, key):