How to create columns from a string in a dataframe?

ndf = (df["long string"]
         .str.extractall(r"(ha|hi|ho):\s\((?:tra|la):\s(\d+)\s(?:tra|la):\s(\d+)\)")
         .droplevel("match")
         .set_index(0, append=True)
         .set_axis(["tra", "la"], axis=1)
         .unstack()
         .swaplevel(axis=1))
ndf.columns = ndf.columns.map("-".join)

• Extract the desired parts with a regex
• Drop the index level induced by extractall called match
• Append the ha-hi-ho matches as the index (0 is first capturing group)
• Rename the remaining columns tra and la
• Unstack the ha-hi-ho index to the columns
• Swap the ha-hi-ho and tra-la levels' order in columns so that ha-hi-ho is upper
• Lastly join these levels of columns' names with a hyphen

to get

  ha-tra hi-tra ho-tra ha-la hi-la ho-la
0      1      1      1     2     2     2
1      1      1      1     2     2     2
2      1      1      1     2     2     2

One way to solve:

df1 = (
    df['long string'].str.extractall(
        r'.*?([a-z]+)\s*?:\s*?\(([a-z]+):\s*(\d+)\s*([a-z]+):\s*(\d+)\)')
    .droplevel("match")
    .set_index(0, append=True)
)

d1 = df1.iloc[:, :2]
d2 = df1.iloc[:, 2:]
d2.columns = d1.columns

df2 = pd.concat([d1, d2]).reset_index()
df2 = df2.pivot(index='level_0', columns=[0, 1], values=2)
df2.columns = df2.columns.map('-'.join)
df2 = df2.reset_index(drop=True)

ALTERNATIVE:

df2 = (
    (
        df['long string'].str.extractall(
            r'.*?([a-z]+)\s*?:\s*?\(([a-z]+):\s*(\d+)\s*([a-z]+):\s*(\d+)\)')
        .droplevel("match")
        .set_index(0, append=True)
        .apply(lambda x: x.values.reshape(-1, 2), axis=1)
        .explode()
        .apply(pd.Series)
        .add_prefix('val')
        .reset_index()
    ).pivot(index=['level_0'], columns=[0, 'val0'], values='val1')
).reset_index(drop=True)
df2.columns = df2.columns.map('-'.join)

OUTPUT:

  ha_la ha_tra hi_la hi_tra ho_la ho_tra
0     2      1     2      1     2      1
1     2      1     2      1     2      1
2     2      1     2      1     2      1