File: //lib/python3/dist-packages/ocrmypdf/__pycache__/hocrtransform.cpython-38.pyc
U
��Z^�6 � @ s6 d dl Z d dlZd dlmZ d dlmZmZmZ d dlm Z d dl
m Z d dl m
Z
eddd d
d g�ZG d d
� d
e�ZG dd� d�Zedk�r2e jdd�Zejdddddd� ejddedd d � ejd d dd d!� ejd"ddd#d� ejd$d%d&� ejd'd(d&� e�� Zeejej�Zej ej ej ej ej d)� dS )*� N)�
namedtuple)�atan�cos�sin)� ElementTree)�inch)�Canvas�Rect�x1�y1�x2�y2c @ s e Zd ZdS )�HocrTransformErrorN)�__name__�
__module__� __qualname__� r r �8/usr/lib/python3/dist-packages/ocrmypdf/hocrtransform.pyr + s r c @ s� e Zd ZdZe�d�Ze�dej�Ze �
dddddd ��Z d
d � Z d d
� Z
dd� Zedd� �Zedd� �Zdd� Zedd� �Zd!dd �Zed d � �Zd d � ZdS )"�
HocrTransformz�
A class for converting documents from the hOCR format.
For details of the hOCR format, see:
http://kba.cloud/hocr-spec/
zbbox((\s+\d+){4})zs
baseline \s+
([\-\+]?\d*\.?\d*) \s+ # +/- decimal float
([\-\+]?\d+) # +/- intZffu ffiu fflZfiZfl)u ffu ffiu fflu fiu flc C s� || _ t�|�| _t�d| j�� j�}d| _|r<|� d�| _d\| _
| _ | j� d| j �D ]8}| �
|�}| �|�}|j|j | _
|j|j | _ q�qZ| j
d ks�| j d kr�td��d S )Nz
({.*})html� � )NNz.//%sdiv[@class='ocr_page']z$hocr file is missing page dimensions)�dpir �parse�hocr�re�matchZgetroot�tag�xmlns�group�width�height�findall�element_coordinates�
pt_from_pixelr r
r
r r )�selfZ hocrFileNamer �matchesZdiv�coordsZ pt_coordsr r r �__init__C s
zHocrTransform.__init__c C s6 | j dkrdS | j �d| j �}|r.| �|�S dS dS )z=
Return the textual content of the HTML body
Nr z .//%sbody)r �findr �_get_element_text)r$ Zbodyr r r �__str__[ s
zHocrTransform.__str__c C sL d}|j dk r||j 7 }|�� D ]}|| �|�7 }q |jdk rH||j7 }|S )zL
Return the textual content of the element and its children
r N)�textZ getchildrenr) �tail)r$ �elementr+ Zchildr r r r) g s
z HocrTransform._get_element_textc sR d}d|j krN| j�|j d �}|rN|�d��� � t�� fdd�td�D ��}|S )zj
Returns a tuple containing the coordinates of the bounding box around
an element
)r r r r �titler c 3 s | ]}t � | �V qd S �N)�int)�.0�n�r& r r � <genexpr> s z4HocrTransform.element_coordinates.<locals>.<genexpr>� )�attrib� box_pattern�searchr �splitr �_make�range)�clsr- �outr% r r3 r r" t s
z!HocrTransform.element_coordinatesc C s@ d|j kr<| j�|j d �}|r<t|�d��t|�d��fS dS )zN
Returns a tuple containing the baseline slope and intercept.
r. r � )r r )r6 �baseline_patternr8 �floatr r0 )r< r- r% r r r �baseline� s
zHocrTransform.baselinec s t �� fdd�|D ��S )zQ
Returns the quantity in PDF units (pt) given quantity in pixels
c 3 s | ]}|� j t V qd S r/ )r r )r1 �c�r$ r r r4 � s z.HocrTransform.pt_from_pixel.<locals>.<genexpr>)r r: )r$ Zpxlr rC r r# � s zHocrTransform.pt_from_pixelc C s |� | j�S )z�
Given an input string, returns the corresponding string that:
- is available in the helvetica facetype
- does not contain any ligature (to allow easy search in the PDF file)
)� translate� ligatures)r< �sr r r �replace_unsupported_chars� s z'HocrTransform.replace_unsupported_charsNF� Helveticac
C s\ t || j| jfdd�}|�ddd� |�ddd� |�d� | j�d| jdf �D ]f}| � |��
� } t | �dkrrqR| � |�}
| �
|
�} |rR|j| j| j| j | j| j | j| j dd� qRd} | j�d| jd f �D ] }
d
} | �||
d ||||� q�| �s&| j�d | jd
f �}| �||d ||||� |dk �rH|j|dd| j| jd� |�� |�� dS )aQ
Creates a PDF file with an image superimposed on top of the text.
Text is positioned according to the bounding box of the lines in
the hOCR file.
The image need not be identical to the image used to create the hOCR
file.
It can have a lower resolution, different color mode, etc.
r )ZpagesizeZpageCompressionr z.//%sp[@class='%s']Zocr_par�ZfillF�.//%sspan[@class='%s']Zocr_lineTZ ocrx_wordz.//%sdiv[@class='%s']Zocr_pageN)r r )r r r �setStrokeColorRGB�setFillColorRGB� setLineWidthr r! r r) �rstrip�lenr" r# �rectr
r
r r �_do_liner( Z drawImageZshowPageZsave)r$ Z outFileNameZ
imageFileName�showBoundingboxes�fontname�
invisibleText�interwordSpaces�pdf�elem�elemtxt�
pxl_coordsZptZ found_lines�line�rootr r r �to_pdf� s^
� ��
�
zHocrTransform.to_pdfc C s ||d |d S )Nr r r )r< Zpoly�xr r r �polyval� s zHocrTransform.polyvalc C sj | � |�}| �|�} | j| j }
| �|�\} } t| �dk r>d} t| �}
t|
�t|
� }}|� � }| | j
t }|
t|� | }|� ||� |r�|�
d� | j| j| }|�r|�� |�ddd� |�d� |�| j|| j| �| |f| j| j �� |�dd� |�dd d � |�|| ||| j|� |�d d d � |�d
| j|f �}|D �]}| �|��� }| � |�}|d k�r|�qP| � |�}| �|�}|�r�|d 7 }t � |j| j|j|� d ||
� | jf�}|j|j }|� |||�}|�r|j |j| j| j ||
d d
� |�!� }|j|d } ||d } |�"| | � |d k�rP|�#d| | � |�$|� �qP|�%|� d S )Ng{�G�zt?g � gffffff�?g�������?g �?� r r rJ r � rI �d )&r" r# r
r rA �absr r r Z beginTextr r ZsetFontZsetTextRenderModer ZsetDashrK rM rZ r
r r^ ZsetTextTransformrL r! r r) �striprG r r: Z stringWidthrP ZgetStartOfLineZ
moveCursorZ
setHorizScaleZtextOutZdrawText) r$ rV rZ Z elemclassrS rT rU rR Zpxl_line_coordsZline_boxZ line_heightZslopeZ
pxl_interceptZangleZcos_aZsin_ar+ Z interceptZfontsizeZ baseline_y2�elementsrW rX rY ZboxZ box_widthZ
font_widthZcursorZdxZdyr r r rQ � s�
�
��
�
zHocrTransform._do_line)NFrH FF)r r r �__doc__r �compiler7 �VERBOSEr? �str� maketransrE r' r* r) � classmethodr" rA r# rG r\ r^ rQ r r r r r / s8
��
�
M
r �__main__zConvert hocr file to PDF)Z descriptionz-bz--boundingboxes�
store_trueFzShow bounding boxes borders)�action�default�helpz-rz --resolutioni, z&Resolution of the image that was OCRed)�typero rp z-iz--imagez-Path to the image to be placed above the text)ro rp z--interword-spaceszAdd spaces between words�hocrfilez"Path to the hocr file to be parsed)rp �
outputfilez$Path to the PDF file to be generated)rU )!�argparser � collectionsr Zmathr r r Z xml.etreer Zreportlab.lib.unitsr Zreportlab.pdfgen.canvasr r � Exceptionr r r �ArgumentParser�parser� add_argumentr0 �
parse_args�argsrr Z
resolutionr r\ rs ZimageZ
boundingboxesZinterword_spacesr r r r �<module> sb 2
�����