From .csv, read only or split into sections separated by "<string>"

Depending on the size of your csv, you could read in the entire file into Pandas and split the dataframe into multiple dataframes via a list comprehension.

data = '''<Network>;;;;;;;;;;;;;;;;;;;;;
            Property;Value;;;;;;;;;;;;;;;;;;;;
            Title;;;;;;;;;;;;;;;;;;;;;
            Version;6.4;;;;;;;;;;;;;;;;;;;;
            ;;;;;;;;;;;;;;;;;;;;;
            <Sites>;;;;;;;;;;;;;;;;;;;;;
            Name;LocationCode;Longitude;Latitude;;;;;;;;;;...'''

df = pd.read_csv(StringIO(data), header=None)

create a list of dataframe names (the headers of each df)

df_names = df[0].str.extract(r'(<[a-zA-Z]+>)')[0].str.strip('<>').dropna().tolist()

find the indices for the headers regions = df.loc[df[0].str.contains(r'<[a-zA-Z]+')].index.tolist()

last_row = df.index[-1]

regions.append(last_row)

from more_itertools import windowed

create windows for each 'sub' dataframe

regions_window = list(windowed(regions,2))

the function helps with some cleanup during the dataframe extraction

def some_cleanup(df):
    df.columns = df.iloc[0].str.extract(r'(<[a-zA-z]+>)')[0].str.strip('<>')
    df = df.iloc[1:]
    return df

extract the dataframes

M = [df.loc[start:end].pipe(some_cleanup) for start,end in regions_window]

create a dict with the keys as the dataframe names

dataframe_dict = dict(zip(df_names,M))

I think you can take simple approach and read txt file like:

with open("dummy.txt") as f:
    lines = f.readlines()

Now just get the location of each section:

sections = [lines.index(line) for line in lines if "<" in line]

Then you can use sections to read in between data in pandas dataframe like:

for i in range(len(sections)):
    header = lines[sections[i]]
    df = pd.DataFrame(lines[sections[i]+1:sections[i+1]],
                      columns=header)
    print(df.head())

There are some great answers here already but I'd recommend a Unix tool! It is shorter and will scale to very large files that don't fit into Pandas.

Assuming your file is called foo.csv:

awk '/< section/{x=i++"foo_mini";next}{print > x;}' foo.csv

Creates as many (numbered) {n}foo_mini.csv files as you have sections. (It seeks the pattern < section, and then starts a new file from the following line.)

Then for completeness' sake, add the csv extension:

for file in *foo_mini; do mv "$file" "${file/foo_mini/foo_mini.csv}"; done

You thus have:

0foo_mini.csv
1foo_mini.csv
etc...

It's then a cinch to read them in with Pandas as separate dataframes, and concat them if you like.

I'd do something like this:

import re
import pandas as pd
new_section = False
header_read = False
data_for_frame = list()
for row in data.splitlines():
    if row.startswith('< '):
        new_section = True
        continue
    if re.match('^\s*$', row):
        new_section = False
        header_read = False
        df = pd.DataFrame(data_for_frame, columns=columns)
        continue
    if new_section:
        if not header_read:
            columns = row.split(' ')
            header_read = True
            continue
        if header_read:
            data_for_frame.append(row.split(' '))
            continue

Import might be only, that the CSV file ends with an empty line as well. And you have to take care about the dataframe naming.

The data.splitlines() just came from my own short test, you have to replace it with with open('myfile','r) as f:and so on.