如果列匹配特定字符串，则在数据框中删除行

Question

I tried to follow the process which was mentioned here but it did not work(completely) for me, so pls point me out to any duplicates which i might be missing, so below is the requirement where i am blocked where I am trying to filter the data with the below condition before inserting data into Postgres.

First name and last name columns should not have [Jr, Sr, I, II, etc] in it. or drop the entire record/row

columns = [
        'cust_last_nm',
        'cust_frst_nm',
        'cust_brth_dt',
        'cust_gendr_cd',
        'cust_postl_cd'
    ]

def push_to_pg_weekly(key):
    total_rows = int(a.split()[0])
    rows = 0
    for chunk in pd.read_csv(key, sep="|", header=None, chunksize=100000):
        rows += len(chunk)
        chunk = chunk.dropna(axis=0)
        chunk = chunk[np.where(
         (chunk[0].astype('str').str.len()>1) & 
         (chunk[1].astype('str').str.len()>1) &
         (chunk[4].astype('str').str.len()>4) &
         (chunk[4].astype('str').str.len()<8), True, False)]
        chunk[0] = ~chunk[0].str.contains("jr", na=False)
        chunk[1] = ~chunk[1].str.contains("jr", na=False)
        chunk.to_csv("/tmp/sample.csv", sep="|", header=None, index=False)
        connection = psycopg2.connect(connection details <here>)
        with connection.cursor() as cursor:
            connection.commit()

Test data that i am working on

jane|doe|1969-01-01|F|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY
jr|doe|1969-01-01|M|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY
jane|sr|1969-01-01|F|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY

i know i am in the right direction, but still missing something else, because when i try this

chunk[0] = ~chunk[0].str.contains("jr", na=False)

i get the below output: instead of False i am expecting that entire row to be dropped

True|True|1969-01-01|F|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY
False|True|1969-01-01|M|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY
True|True|1969-01-01|F|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY

Expected Output:

jane|doe|1969-01-01|F|926.0|1351127|E2sboFz4Mk2aGIKhD4vm6J9Jt3ZSoSdLm+0PCdWsJto=|YSILMFS5sPPZZF/KFroEHV77z1bMeiL/f4FqF2kj4Xc=|tNjgnby5zDbfT2SLsCCwhNBxobSDcCp7ws0zYVme5w4=|kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=|cigna_TOKEN_ENCRYPTION_KEY

And another question i have is: can i try including multiple parameters in str.contains to filter more conditions, i tried the below two methods but it did not work, both of them yielded the true/false results as well

chunk[0] = ~chunk[0].str.contains("jr", “sr”, “|”, “||”, na=False)
chunk[1] = ~chunk[1].str.contains("jr", “sr”, “|”, “||”,  na=False)
or
searchfor = [‘jr’, ’sr’,’|’,’||’]
chunk[0] = ~chunk.chunk[0].str.contains('|'.join(searchfor))]
chunk[1] = ~chunk.chunk[1].str.contains('|'.join(searchfor))]

Or should i be using drop method to drop rows, any suggestions or comments will be appreciated, thanks

Answer 1

Essentially you are forgetting to pass the boolean series (True/False) into brackets [...] or better with .loc[...] . Instead, you are re-assigning the values within those chunk columns to the result of your conditions but not applying conditions logically to the data frame.

Therefore, consider calling .loc[] with intersection of both those conditions:

# ASSIGN BOOLEAN SERIES
fname_jr = ~chunk.loc[0].str.contains("jr", na=False)
lname_jr = ~chunk.loc[1].str.contains("jr", na=False)

# PASS INTO .loc
chunk_sub = chunk.loc[fname_jr & lname_jr]
chunk_sub

#       0    1   ...                                            9                          10
# 0  jane  doe  ...  kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=  cigna_TOKEN_ENCRYPTION_KEY
# 2  jane   sr  ...  kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=  cigna_TOKEN_ENCRYPTION_KEY

And to integrate multiple selections, call str.join to combine a list of items with pipe-delimiters:

# ASSIGN BOOLEAN SERIES
fname_jr_sr = ~chunk[0].str.contains("|".join(["sr", "jr"]), na=False)
lname_jr_sr = ~chunk[1].str.contains("|".join(["sr", "jr"]), na=False)

# PASS INTO .loc
chunk_sub = chunk.loc[fname_jr_sr & lname_jr_sr]
chunk_sub
#       0    1   ...                                            9                          10
# 0  jane  doe  ...  kk25p0lrp2T54Z3B1HM3ZQN0RM63rjqvewrwW5VhYcI=  cigna_TOKEN_ENCRYPTION_KEY

---

Relatedly, your np.where call is not necessary as .loc will run on boolean series. Be sure to also escape | with backslashes \\\\ since the pipe symbol is a string matching operator. Altogether:

chunk = chunk.loc[(chunk[0].astype('str').str.len()>1) & 
                  (chunk[1].astype('str').str.len()>1) &
                  (chunk[4].astype('str').str.len()>4) &
                  (chunk[4].astype('str').str.len()<8) & 
                  ~chunk[0].str.contains("|".join(["sr", "jr", "\\|", "\\|\\|"]), na=False) & 
                  ~chunk[1].str.contains("|".join(["sr", "jr", "\\|", "\\|\\|"]), na=False)]

chunk.to_csv("/tmp/sample.csv", sep="|", header=None, index=False)