4
$\begingroup$

I have an html file like this:

<h1>Group 1</h1>
<table>
  <tr>
    <td>Col1</td>
    <td>Col2</td>
    <td>Col3</td>
  </tr>
  <tr>
    <td>ValA</td>
    <td>ValB</td>
    <td>ValC</td>
  </tr>
</table>
<h1>Group 2</h1>
<table>
  <tr>
    <td>Col1</td>
    <td>Col2</td>
    <td>Col3</td>
  </tr>
  <tr>
    <td>ValP</td>
    <td>ValQ</td>
    <td>ValR</td>
  </tr>
</table>

I want to read it into Pandas as if it had this structure:

<table>
 <tr>
   <td>Caption</td>
   <td>Col1</td>
   <td>Col2</td>
   <td>Col3</td>
 </tr>
 <tr>
   <td>Group 1</td>
   <td>ValA</td>
   <td>ValB</td>
   <td>ValC</td>
 </tr>
 <tr>
   <td>Group 2</td>
   <td>ValP</td>
   <td>ValQ</td>
   <td>ValR</td>
 </tr>
</table>

I can do it easily with PowerQuery the language for PowerBI:

let
    Source = Web.Page(File.Contents("multiple_tables.html")),
    #"Expanded Data" = Table.ExpandTableColumn(Source, "Data", {"Column1", "Column2", "Column3"}, {"Col1", "Col2", "Col3"}),
    #"Filtered Rows" = Table.SelectRows(#"Expanded Data", each ([Caption] <> "Document") and ([Col1] <> "Col1"))
in
    #"Filtered Rows"

Is there a way to achieve this effect in less than 10 lines of code using Python/Pandas plus some html parser opensource library? Or should I resign myself to writing lower level code to handle this?

$\endgroup$

1 Answer 1

3
$\begingroup$

With a bit of parsing via BeautifulSoup, we can get a pandas.Dataframe using pandas.read_html() like:

Code:

def get_tables(source):
    elems = iter(BeautifulSoup(source, 'lxml').find_all(['table', 'h1']))
    df = pd.DataFrame(
        pd.read_html(str(next(elems)), header=0)[0].iloc[0].rename(h1.text)
        for h1 in elems)
    df.index.names = ['Caption']
    return df

Test Code:

import pandas as pd
from bs4 import BeautifulSoup

with open('test.html', 'r') as f:
    print(get_tables(f))

Results:

         Col1  Col2  Col3
Caption                  
Group 1  ValA  ValB  ValC
Group 2  ValP  ValQ  ValR
$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.