Я думаю, что это гораздо сложнее, потому что некоторые столбцы плохо разбирается read_fwf
и некоторые пост-обработка необходима для колонн 3
- 5
в df
cols1
и колонки 8
к df
cols2
с функциями str.split
, shift
, iloc
и drop
. Затем используйте concat
для объединения всех вместе:
import pandas as pd
import io
temp=u"""<TABLE>
<CAPTION>
FORM 13F INFORMATION TABLE
COLUMN 1 COLUMN 2 COLUMN 3 COLUMN 4 COLUMN 5 COLUMN 6 COLUMN 7 COLUMN 8
---------------------------- ---------------- --------- ----------- ------------------- ---------- -------- ----------------------
VALUE SHRS OR SH/ PUT/ INVESTMENT OTHER VOTING AUTHORITY
NAME OF ISSUER TITLE OF CLASS CUSIP (x$1000) PRN AMT PRN CALL DISCRETION MANAGERS SOLE SHARED NONE
---------------------------- ---------------- --------- ----------- ---------- --- ---- ---------- -------- ---------- ------ ----
<S> <C> <C> <C> <C> <C> <C> <C> <C> <C> <C> <C>
7 DAYS GROUP HLDGS LTD ADR 81783J101 19,317 999,322 SH SOLE 999,322 0 0
ACCENTURE PLC IRELAND SHS CLASS A G1151C101 200,952 3,325,917 SH SOLE 3,325,917 0 0
ACCRETIVE HEALTH INC COM 00438V103 85,394 2,966,088 SH SOLE 2,966,088 0 0"""
#after testing replace io.StringIO(temp) to filename
df = pd.read_fwf(io.StringIO(temp), skiprows=[0,1,2,3,5,8,9])
print df
COLUMN 1 COLUMN 2 \
0 NaN NaN
1 NAME OF ISSUER TITLE OF CLASS
2 7 DAYS GROUP HLDGS LTD ADR
3 ACCENTURE PLC IRELAND SHS CLASS A
4 ACCRETIVE HEALTH INC COM
COLUMN 3 COLUMN 4 COLUMN 5 COLUMN 6 COLUMN 7 \
0 VALUE SHRS OR SH/ PUT/ INVESTMENT OTHER
1 CUSIP (x$1000) PRN AMT PRN CALL DISCRETION MANAGERS
2 81783J101 19,317 999,322 SH SOLE NaN
3 G1151C101 200,952 3,325,917 SH SOLE NaN
4 00438V103 85,394 2,966,088 SH SOLE NaN
COLUMN 8
0 VOTING AUTHORITY
1 SOLE SHARED NONE
2 999,322 0 0
3 3,325,917 0 0
4 2,966,088 0 0
#split columns and create new df
cols1 = df.iloc[:, 2].str.split(expand=True)
#shift first row
cols1.iloc[0,:] = cols1.iloc[0,:].shift()
#concanecate columns
cols1.iloc[[0,1], 2] = cols1.iloc[[0,1], 2] + ' ' + cols1.iloc[[0,1], 3]
cols1.iloc[[0,1], 3] = cols1.iloc[[0,1], 4]
#remove column 4
cols1 = cols1.drop(4, axis=1)
#replace , to empty string with 1. and 2. columns
cols1.iloc[2:,1] = cols1.iloc[2:,1].str.replace(',', '')
cols1.iloc[2:,2] = cols1.iloc[2:,2].str.replace(',', '')
print cols1
0 1 2 3 5
0 NaN VALUE SHRS OR SH/ PUT/
1 CUSIP (x$1000) PRN AMT PRN CALL
2 81783J101 19317 999322 SH None
3 G1151C101 200952 3325917 SH None
4 00438V103 85394 2966088 SH None
#split columns and create new df
cols2 = df.iloc[:, 5].str.split(expand=True)
#replace , to empty string
cols2.iloc[2:,0] = cols2.iloc[2:,0].str.replace(',', '')
print cols2
0 1 2
0 VOTING AUTHORITY None
1 SOLE SHARED NONE
2 999322 0 0
3 3325917 0 0
4 2966088 0 0
df = pd.concat([df.iloc[:,[0,1]], cols1, df.iloc[:,[3,4]], cols2], axis=1)
df.columns = range(12)
print df
0 1 2 3 4 5 \
0 NaN NaN NaN VALUE SHRS OR SH/
1 NAME OF ISSUER TITLE OF CLASS CUSIP (x$1000) PRN AMT PRN
2 7 DAYS GROUP HLDGS LTD ADR 81783J101 19317 999322 SH
3 ACCENTURE PLC IRELAND SHS CLASS A G1151C101 200952 3325917 SH
4 ACCRETIVE HEALTH INC COM 00438V103 85394 2966088 SH
6 7 8 9 10 11
0 PUT/ INVESTMENT OTHER VOTING AUTHORITY None
1 CALL DISCRETION MANAGERS SOLE SHARED NONE
2 None SOLE NaN 999322 0 0
3 None SOLE NaN 3325917 0 0
4 None SOLE NaN 2966088 0 0
Если вам нужны столбцы имена из ряда 1
и 2
использования reset_index
, а затем преобразовать строки столбцы to_numeric
:
#column names from 2 rows to 1
df.iloc[1, 3:11] = df.iloc[0, 3:11] + ' ' + df.iloc[1, 3:11]
df.columns = df.iloc[1,:]
#data are from 2 rows (1,2 rows is header)
df1 = df.iloc[2:,:].reset_index(drop=True)
df1.columns.name = None
df1.iloc[:, 3] = pd.to_numeric(df1.iloc[:, 3])
df1.iloc[:, 4] = pd.to_numeric(df1.iloc[:, 4])
df1.iloc[:, 9] = pd.to_numeric(df1.iloc[:, 9])
df1.iloc[:, 10] = pd.to_numeric(df1.iloc[:, 10])
print df1
NAME OF ISSUER TITLE OF CLASS CUSIP VALUE (x$1000) \
0 7 DAYS GROUP HLDGS LTD ADR 81783J101 19317
1 ACCENTURE PLC IRELAND SHS CLASS A G1151C101 200952
2 ACCRETIVE HEALTH INC COM 00438V103 85394
SHRS OR PRN AMT SH/ PRN PUT/ CALL INVESTMENT DISCRETION OTHER MANAGERS \
0 999322 SH None SOLE NaN
1 3325917 SH None SOLE NaN
2 2966088 SH None SOLE NaN
VOTING SOLE AUTHORITY SHARED NONE
0 999322 0 0
1 3325917 0 0
2 2966088 0 0
print df1.dtypes
NAME OF ISSUER object
TITLE OF CLASS object
CUSIP object
VALUE (x$1000) int64
SHRS OR PRN AMT int64
SH/ PRN object
PUT/ CALL object
INVESTMENT DISCRETION object
OTHER MANAGERS object
VOTING SOLE int64
AUTHORITY SHARED int64
NONE object
dtype: object
Как это работает? – jezrael
он выглядит хорошо. спасибо, что показал мне, как – jason