df = pd.DataFrame(para, columns=('interval_hour', 'interval_sec', 'pre','tem', 'WIN_S_Avg_10mi', 'WIN_D_Avg_10mi', 'RHU', 'pwv', 'ztd', 'PRE_1h'))

# reunit data
df_merge = pd.merge(df_none_PRE_1h, df_arc, on="interval_hour", how="left")
# 按不同列标题合并
df_merge = pd.merge(ChinaMetList, igraList,left_on='id',right_on='wmo', how="left")
# throw mismatching epoch out
df_merge = df_merge.dropna(axis=0, how='any')#按多个键合并
df_merge = pd.merge(ztd_panda, ztd_qxj, on=['site', 'date'], how="left")

merge3 = merge3.dropna(subset=['id_cmonoc']).reset_index(drop=True)
df = df.dropna(subset=('pressure', 'temperature', 'dewpoint'), how='any').reset_index(drop=True)

header = df.columns.tolist()

array = np.array(df)

DataFrame.copy(deep=True)

igraSites = pd.DataFrame(columns=header)
igraSites.loc[len(igraSites)] = site     //site是一个list//loc表示按索引获取行，因为df索引从0开始，len表示新的一行

igraSites[['lat', 'lon', 'alt']] = igraSites[['lat', 'lon', 'alt']].astype('float')

minIndex = igraSites['distance'].idxmin()  //查找索引
minRow = igraSites.loc[minIndex]    //获取行

df = pd.read_table(path, sep='\t', header=None)
//sep 表示分隔符，\t表制表符，一般为换行符，存在于行尾
//header=None 表示第一行不作为表头
//

import pandas as pd
header = list(range(10))
df = pd.read_table('cmonocSits.txt' ,header=None, names = header,delim_whitespace=True)
#names 表示列标题
#假如列过多不清楚header中列标题的数量时，可以先用下列方式查看，再添加列标题，注意新列名的长度必须与旧列名一致
out = pd.read_table(fpath, header=None, delim_whitespace=True)
print(out .columns)header = ['site', 'mjd', 'total year', 'year_doy_time', 'ztd', 'nepo', 'flag','nobs']out = pd.read_table(fpath, header=None, names=header,dtype={'year_doy_time': str},skiprows=3,delim_whitespace=True)out['year_doy_time'] = pd.to_datetime(out['year_doy_time'], format='%Y.%j%H%M')# dtype 指定导入列的类型
# skiprows 跳过的行数
# pd.to_datetime处理时间字符串

# csv
df.to_csv( 'igra2-station-list.csv',index=None)
igraSites = pd.read_csv('igra2-station-list.csv')# txt
df_cma_fillter.to_csv('df_cma.txt',sep='\t', index=None,header=0)# xlsx
df = pd.read_excel(fpath, header=None, names=header, index=False, skiprows=headnum)

#遍历行
for index, row in df.iterrows():print (row["lat_x"], row["lat_y"])#遍历列
for index, item in commonSet.iteritems():print(item )

#新增单列
df['c1'] = None
#新增多列
df[['c1','c2']] == None  #报错 KeyError: "['c1','c2'] not in index"
使用concat
pd.concat([df, pd.DataFrame(columns=['c1','c2'])]) #该方法不能只能增加列的位置
使用reindex来重排和增加列名
df.reindex(columns=['c1','c2'])   #删除列
del curr_data['date']

    df_merge.loc[index,['sitename', 'lat_z', 'lon_z', 'distance']] = minIndex,minLat,minLon,minDis

df.index = range(len(df))

out = out.set_index('列名')

curr_df = inf.to_frame().stack().unstack(level=0)

curr_df = curr_df.append([row]*5) 

curr_df['date'] = pd.to_datetime(curr_df[['year','month','day','hour']])
#curr_df表示时间年月日时的列标题必须为'year','month','day','hour'的小写或混合大小写

 curr_df['date'] = curr_df['date_bj'] - pd.Timedelta(8, unit='h')

match['obs'] = (match['precipitation'] >= typeThreshold).astype('int')
//
match['obs'][match['obs']>= typeThreshold]=1
match['obs'][match['obs']< typeThreshold]=0
//
match['obs'].mask(match['obs'] >=  typeThreshold, 1, inplace=True)
match['obs'].mask(match['obs'] < typeThreshold, 0, inplace=True)