Want to extract information by separate two parts

[youpengbo2018]

[China_ptent.pdf](https://github.com/jsvine/pdfplumber/files/7750238/Chin
a_ptent.pdf)

Hi, I wonder that extract the file's data by split into two parts based on the picture.
.This is the code that I try to extract those information before.
import pdfplumber as pr
import pandas as pd
filename = r'F:\GB2652.pdf'
pdf=pr.open(filename)
page = pdf.pages[0]
table = page.extract_tables(table_settings={
"vertical_strategy":"text",
"horizontal_strategy":"text",
"join_tolerance":50
Thank you for your help

[youpengbo2018]

China_ptent.pdf

[jsvine]

Hi @youpengbo2018, and thanks for your interest in this library. Rather than treating the information as tables, you may have better luck with a more customized approach. Luckily, the PDFs appear to have an obvious vertical line (see page.lines), and then the ----------------- marks as horizontal lines, separating the portions of each page. I would try identifying those, and then using page.crop(...) to select each section that you calculate from that.

To get the horizontal dividers, try:

import re
horizontal_dividers = [ w for w in page.extract_words()
  if re.search(r"^-{10,}$", w["text"]) ]

[youpengbo2018]

Thank you for your help. Could you describe more details about this, because I have not use crop to extract information. In addition, is that possible I can still use the extract_table to extact the information. I use these code to achieve my goals that to extract the information and store in a dataframe. Now the only challenge that I have faced that is sometimes I use the table_extract method, the methods can not get the last row with the (--------------）mark such as the page 11 of the China_ptent file

china-patent.pdf

These are the codes I have used to achieve the goal.

from decimal import Decimal
import pdfplumber as pr
import re
import pandas as pd
import time

def get_table(pagenum, pdf):
    page = pdf.pages[pagenum]
    table = page.extract_tables(table_settings={
        "vertical_strategy":"text",
        "horizontal_strategy":"text",
        "join_tolerance":50
    })
    return table


def get_page_information2():
    filename = r'F:\FMGKGB2652.pdf'
    pdf = pr.open(filename)
    lastline = ""
    for i in range(9,12):
    # for i in range(len(pdf.pages)):
        print(i)
        table2 = get_table(i, pdf)[0]
        list1 = []
        list2 = []
        for j in range(len(table2)-1):
            list1.append(table2[j][0])
            try:
                list2.append(table2[j][1])
            except:
                pass

            leftpage = ''.join(list1)
            rightpage = ''.join(list2)
        lines = (lastline + leftpage + rightpage).split('-----------------------------------------')
        # print("==========================================lines",lines)
        print(rightpage[-100:])
        for k, line in enumerate(lines):
            # print("enum",k,line)
            line = line.replace("\r\n", "").replace("\n", "").replace('\r', "")
            if k < len(lines) - 1:
                yield line
            else:
                lastline = line

def extract_value(pattern, value):
    try:
        return re.findall(pattern, value)[0][:-4]
    except:
        return None
        pass

def extract_value_last(pattern, value):
    try:
        return re.findall(pattern, value)[0]
    except:
        return None
        pass
def get_var_value(value):       # extract different column value by extract_value method
        valuelist=['v10','v51','v22','v21','v43','v71','v72','v74','v54','v57']
        list8=[]
        pattern10=r'\(10\).*?\(\d\d\)'
        pattern51=r'\(51\).*?\(\d\d\)'
        pattern21=r'\(21\).*?\(\d\d\)'
        pattern22=r'\(22\).*?\(\d\d\)'
        pattern71=r'\(71\).*?\(\d\d\)'
        pattern72=r'\(72\).*?\(\d\d\)'
        pattern54=r'\(54\).*?\(\d\d\)'
        pattern57=r'\(57\).*'
        pattern43 = r'\(43\).*?\(\d\d\)'
        pattern74 = r'\(74\).*?\(\d\d\)'
        v10=extract_value(pattern10,value)
        v51=extract_value(pattern51,value)
        v21 = extract_value(pattern21, value)
        v22 = extract_value(pattern22, value)
        v71 = extract_value(pattern71, value)
        v72 = extract_value(pattern72,value)
        v43 = extract_value(pattern43, value)
        v74 = extract_value(pattern74, value)
        v54 = extract_value(pattern54,value)
        v57 = extract_value_last(pattern57,value)
        list8.append(v51)
        list8.append(v10)
        list8.append(v21)
        list8.append(v22)
        list8.append(v43)
        list8.append(v71)
        list8.append(v72)
        list8.append(v74)
        list8.append(v54)
        list8.append(v57)
        return list8


if __name__ == '__main__':
    list7=[]
    now=time.localtime()
    i=0
    for item in get_page_information2():++++++++++++++++++++++++++++++++++++++record：",i, item)
        i+=1
        list7.append(get_var_value(item))

    df1=pd.DataFrame(list7)
    df1.to_csv('F:\\patenttest.csv')
    # print(df1)
    print(time.localtime())
    print(now)

[jsvine]

Hi @youpengbo2018, here's a notebook that demonstrates what I was talking about — hopefully this clarifies things: https://notebooksharing.space/view/3cf2635360372719bc95152b027cdfcdb7938658d1575b90d6450399870bcfd1

Since the PDF doesn't really have proper tables, .extract_tables(...) will probably not get you the results you desire.

[youpengbo2018]

Hi,thank you for the help. It really help me a lot.