大熊貓和大型數據框

Question

我決定使用熊貓（0.18.1）從我的一個模型中使用離散元素粒子處理日誌數據。該日誌具有與400000個顆粒（X，Y，Z的位置和速度;圍繞5M行）屬性結構如下：

***************************************** 
* Log File Started 16:12:54 Fri May 06 2016 
* 4.00-182 (64-bit) 
* 
* 
*  
***************************************** 
elrond> 
Ball_Id 400000 
Ballx 4.90707890560e+002 
Bally 9.19154644947e+001 
Ballz -1.02229145082e+002 
Top 0 
Dx 1.38904597749e+000 
Dy -6.35282219552e-001 
Dz -1.64199872399e+001 
Velx -1.02171891554e-001 
Vely -1.05325799073e-002 
Velz 4.04701964190e-003 
V_rotx -6.86579713474e-004 
V_roty 9.14539972137e-004 
V_rotz -7.76239471255e-005 
Ball_Id 399999 
Ballx 7.48469370428e+002 
Bally 2.46351257548e+001 
Ballz -8.62490399310e+001 
Top 0 
Dx 6.96274451933e-001 
Dy 1.32036797483e+000 
Dz -1.87517847236e+001 
Velx -1.05970416552e-002 
Vely 7.21491947832e-003 
Velz 7.55093644847e-004 
V_rotx 5.17377621567e-006 
V_roty 2.59041151397e-005 
V_rotz -2.31863427848e-005 
Ball_Id 399998 
Ballx 1.19395239848e+002 
Bally 7.80444921824e+001 
Ballz 2.34352803814e+000 
Top 0 
Dx 5.90917177795e+001 
Dy 1.37004693793e+000 
Dz 1.61822040639e+001 
Velx 1.31243808962e+001 
Vely -8.20542806383e-001 
Velz 6.19737823128e+000 
V_rotx -4.89777825136e-002 
V_roty 9.36324827264e-002 
V_rotz -5.90727285357e-002 

我想用這種格式的文件：

Ball_Id Ballx Bally Ballz Topo Dx Dy Dz Velx Vely Velz V_rotx V_roty V_rotz 
400000 4.90714073236e+002 9.19065373175e+001 -1.02231392317e+002 0 1.39522865407e+000 -6.44209396797e-001 -1.64222344741e+001 2.68881171417e-002 -1.81227520077e-002 -4.04738585013e-003 7.75669240314e-005 -4.00875407555e-004 -1.41810083383e-004 
399999 7.48472521138e+002 2.46451444724e+001 -8.62470162686e+001 0 6.99425161310e-001 1.33038669240e+000 -1.87497610612e+001 1.18932839949e-002 4.69256261481e-003 1.38621378252e-002 -6.30154171502e-006 -3.23043526114e-004 2.16368702869e-007 
399998 1.28116171848e+002 7.67039376593e+001 7.55623907648e+000 0 6.78126497794e+001 2.94924148016e-002 2.13949151023e+001 6.33940244884e+000 1.73376959946e-001 4.85967665797e+000 -3.52816583310e-001 -5.38872247688e-001 1.12736371677e-001 
399996 4.79841096924e+002 -1.62882386399e+002 -1.30791611129e+002 Topo1 2.73837679243e+000 -1.47077675894e+000 -6.28235946603e+000 7.90493795999e-002 -3.39089755154e-002 1.02726075741e-003 -1.14738159279e-004 -7.24753898272e-005 -6.78627383629e-005 

到目前爲止，我能夠編寫一個非常低效的代碼，需要一個永恆才能達到我想要的最終文件。任何改善它的建議都會很棒。 感謝

import pandas as pd 
#================================================================================= 
df = pd.read_csv("Desloc_Caixa_Compress_14_04_16_19.log",index_col=0,header = None, skiprows =[0,1,2,3,4,5,6,7],engine='python',skipfooter = 4, sep=" ") 
dados = df[0:14] 
#================================================================================= 
k=14; f=28; m=28; n=42 
while (n<=len(df)): 
    a=df[k:f] 
    b=df[m:n] 
    k+=28; f+=28 
    m+=28; n+=28 
    dados = pd.concat([dados,a, b], axis=1) 
#=================================================================================  
d= dados.transpose() 
data = d.set_index('Ball_Id') 
data.to_csv('Data_14_04_16_19.txt', sep='\t') 
#================================================================================= 

Answer 1

你可以使用df.pivot：

import pandas as pd 
df = pd.read_csv("Desloc_Caixa_Compress_14_04_16_19.log", header=None, 
       skiprows=8, engine='python', skipfooter=4, sep=" ") 

df['index'] = (df[0] == 'Ball_Id').cumsum() 
df = df.pivot(index='index', columns=0, values=1) 

產生

0  Ball_Id  Ballx  Bally  Ballz   Dx  Dy 
index                  
1  400000.0 490.707891 91.915464 -102.229145 1.389046 -0.635282 
2  399999.0 748.469370 24.635126 -86.249040 0.696274 1.320368 
3  399998.0 119.395240 78.044492 2.343528 59.091718 1.370047 

                      \ 
0    Dz Top V_rotx V_roty V_rotz  Velx  Vely 
index                  
1  -16.419987 0.0 -0.000687 0.000915 -0.000078 -0.102172 -0.010533 
2  -18.751785 0.0 0.000005 0.000026 -0.000023 -0.010597 0.007215 
3  16.182204 0.0  NaN  NaN  NaN 13.124381 -0.820543 


0   Velz 
index    
1  0.004047 
2  0.000755 
3   NaN