Pandas Dataframes - Is there a better way to add rows and then append some data to those rows?

user_18784484 0
Viewed 67

I haven't seen something asked like this, so I write some code like:

import pandas as pd
import pyodbc
from collections import Counter
import itertools

server = ""
cnxn = ""
query = ("")
try:
    df = pd.read_sql(query, cnxn).astype('string')
except:
    print("query failed")
else:
    cnxn.close()

which returns a dataframe like this:

partNo    planStatus    planRev    operation
------    ----------    -------    ---------
110068    Released      2A         0100-00-0
110068    Released      2A         0200-00-0
110383    Released      3B         0100-00-0
110383    Released      3B         0200-00-0
110384    In Dev        1C         0100-00-0
110384    In Dev        1C         0200-00-0

so now I want to add a row for each part number that has an operation '000':

dfNums= list(df['partNo'].drop_duplicates())
temp = pd.DataFrame({'partNo':dfNums, 'operation':['000']*len(dfNums)})
df = pd.concat([df, temp]).sort_values(by=['partNo', 'operation'])

which returns a dataframe like this:

partNo    planStatus    planRev    operation
------    ----------    -------    ---------
110068                             000
110068    Released      2A         0100-00-0
110068    Released      2A         0200-00-0
110383                             000
110383    Released      3B         0100-00-0
110383    Released      3B         0200-00-0
110384                             000
110384    In Dev        1C         0100-00-0
110384    In Dev        1C         0200-00-0

So now in order to get planStatus and planRev to populate to the '000' operation row, the best way I could think of was:

for num in dfNums:
    getNumRevs = list(df.loc[df['partNo'] == num]['planRev'])
    getNumStatus = list(df.loc[df['partNo'] == num]['planStatus'])
    data = Counter(getNumRevs)
    data1 = Counter(getNumStatus)
    mostCommonRev = max(getNumRevs, key=data.get)
    mostCommonStatus = max(getNumStatus, key=data1.get)
    df.loc[df['partNo'] == num, 'planRev'] = mostCommonRev
    df.loc[df['partNo'] == num, 'planStatus'] = ""
    df.loc[(df['partNo'] == num) & (df['operation'] == '000'), 'planStatus'] = mostCommonStatus

Which I can't imagine is anywhere near the most efficient way to do this. Is there a better way to do this using groupby? Or is there just any better way to do this? This just gives me a gross feeling of iterating through dataframes but it's the only way I could achieve the output I want, which looks like:

partNo    planStatus    planRev    operation
------    ----------    -------    ---------
110068    Released      2A         000
110068                  2A         0100-00-0
110068                  2A         0200-00-0
110383    Released      3B         000
110383                  3B         0100-00-0
110383                  3B         0200-00-0
110384    In Dev        1C         000
110384                  1C         0100-00-0
110384                  1C         0200-00-0

Edit for @rayad's comment: This is similar to what @rayad's comment was leading to... I did something relatively similar that seemed faster, I never did test the speed though.

# create a temp df of all part nums and operation 000
routingNums = list(df['partNo'].drop_duplicates())
temp = pd.DataFrame({'partNo':routingNums, 'operation':['000']*len(routingNums)})
# add the temp df of '000' ops to the main df and sort
df = pd.concat([df, temp]).sort_values(by=['partNo', 'operation']).reset_index(drop=True)
# make all cimxDatabase values WLCAPP
df.loc[df['operation'] == '000', 'cimxDatabase'] = 'WLCAPP'
# make all '000' op rows have the same planRev and planStatus as the rest of the partNo's associated
check = df[['partNo','planRev','planStatus']].drop_duplicates(subset='partNo', keep='last')
df_to_merge = df[['partNo']].merge(check, on='partNo', how='left')
df.update(df_to_merge, overwrite=True)
pandas dataframe python
1 Answers

With the first dataframe you provided:

import pandas as pd

df = pd.DataFrame([
    {"partNo": 110068,"planStatus": "Released", "planRev": "2A", "operation": "0100-00-0",},
    {"partNo": 110068, "planStatus": "Released", "planRev": "2A", "operation": "0200-00-0",},
    {"partNo": 110383, "planStatus": "Released", "planRev": "3B", "operation": "0100-00-0",},
    {"partNo": 110383, "planStatus": "Released", "planRev": "3B", "operation": "0200-00-0",},
    {"partNo": 110384, "planStatus": "In Dev", "planRev": "1C", "operation": "0100-00-0",},
    {"partNo": 110384, "planStatus": "In Dev", "planRev": "1C", "operation": "0200-00-0",},
    ])

Here is another way to do it:

# Create and add new rows
new_rows = [
    pd.DataFrame(
        {
            "partNo": [partno],
            "planStatus": [pd.NA],
            "planRev": [pd.NA],
            "operation": ["000"],
        }
    )
    for partno in df["partNo"].unique()
]

df = (
    pd.concat([df, *new_rows])
    .sort_values(by=["partNo", "operation"])
    .fillna(method="bfill")
    .reset_index(drop=True)
)

# Remove duplicated values in `planStatus` column
df.loc[df["operation"] != "000", "planStatus"] = ""

And so:

print(df)
# Output
   partNo planStatus planRev  operation
0  110068   Released      2A        000
1  110068                 2A  0100-00-0
2  110068                 2A  0200-00-0
3  110383   Released      3B        000
4  110383                 3B  0100-00-0
5  110383                 3B  0200-00-0
6  110384     In Dev      1C        000
7  110384                 1C  0100-00-0
8  110384                 1C  0200-00-0
user_112460560
Related