调用 TabPy SCRIPT_REAL 时出现 Tableau 错误“所有字段必须是聚合的或常量”

Question

我正在通过 Tableau 工作表中的计算字段调用 TabPy 服务器以运行假设检验：预订率是否因组而异？

我有一张桌子，例如：

     Group  Bookings
0        A         1
1        A         0
3998     B         1
3999     B         0

在 Python 中，在同一台服务器 (using the python 2.7 docker image) 上，我想要的测试很简单：

from scipy.stats import fisher_exact
df_cont_tbl = pd.crosstab(df['Group'], df['Bookings'])
prop_test = fisher_exact(df_cont_tbl)
print 'Fisher exact test: Odds ratio = {:.2f}, p-value = {:.3f}'.format(*prop_test)

返回：Fisher exact test: Odds ratio = 1.21, p-value = 0.102

我将 Tableau 连接到 TabPy 服务器，并且可以执行 hello-world 计算字段。例如，我用计算字段返回 42：SCRIPT_REAL("return 42", ATTR([Group]),ATTR([Bookings]) )

但是，我尝试使用计算字段调用上面的 stats 函数来提取 p 值：

SCRIPT_REAL("
import pandas as pd
from scipy.stats import fisher_exact
df_cont_tbl = pd.crosstab(_arg1, _arg2)
prop_test = fisher_exact(df_cont_tbl)
return prop_test[1]
", [Group], [Bookings] )

我收到通知：计算包含错误下拉列表使用表计算函数或来自多个数据源的字段时，所有字段必须是聚合或常量

我尝试用ATTR() 包装输入，如下所示：

SCRIPT_REAL("
import pandas as pd
from scipy.stats import fisher_exact
df_cont_tbl = pd.crosstab(_arg1, _arg2)
prop_test = fisher_exact(df_cont_tbl)
return prop_test[1]
",ATTR([Group]), ATTR([Bookings])
)

这会将通知更改为“计算有效”但从服务器返回 Pandas ValueError：

An error occurred while communicating with the External Service.
Error processing script
Error when POST /evaluate: Traceback
Traceback (most recent call last):
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/tabpy_server/tabpy.py", line 467, in post
result = yield self.call_subprocess(function_to_evaluate, arguments)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/tornado/gen.py", line 1008, in run
value = future.result()
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/tornado/concurrent.py", line 232, in result
raise_exc_info(self._exc_info)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/tornado/gen.py", line 1014, in run
yielded = self.gen.throw(*exc_info)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/tabpy_server/tabpy.py", line 488, in call_subprocess
ret = yield future
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/tornado/gen.py", line 1008, in run
value = future.result()
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/concurrent/futures/_base.py", line 400, in result
return self.__get_result()
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/concurrent/futures/_base.py", line 359, in __get_result
reraise(self._exception, self._traceback)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/concurrent/futures/_compat.py", line 107, in reraise
exec('raise exc_type, exc_value, traceback', {}, locals_)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/concurrent/futures/thread.py", line 61, in run
result = self.fn(*self.args, **self.kwargs)
File "<string>", line 5, in _user_script
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/pandas/tools/pivot.py", line 479, in crosstab
df = DataFrame(data)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/pandas/core/frame.py", line 266, in __init__
mgr = self._init_dict(data, index, columns, dtype=dtype)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/pandas/core/frame.py", line 402, in _init_dict
return _arrays_to_mgr(arrays, data_names, index, columns, dtype=dtype)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/pandas/core/frame.py", line 5398, in _arrays_to_mgr
index = extract_index(arrays)
File "/opt/conda/envs/Tableau-Python-Server/lib/python2.7/site-packages/pandas/core/frame.py", line 5437, in extract_index
raise ValueError('If using all scalar values, you must pass'
ValueError: If using all scalar values, you must pass an index
Error type : ValueError
Error message : If using all scalar values, you must pass an index

示例数据集：

要生成我要连接的 CSV：

import os
import pandas as pd
import numpy as np
from collections import namedtuple

OUTPUT_LOC = os.path.expanduser('~/TabPy_demo/ab_test_demo_results.csv')

GroupObs = namedtuple('GroupObs', ['name','n','p'])

obs = [GroupObs('A',3000,.10),GroupObs('B',1000,.13)] 
# note true odds ratio = (13/87)/(10/90) = 1.345

np.random.seed(2019)

df = pd.concat( [ pd.DataFrame({'Group': grp.name,
                                'Bookings':  pd.Series(np.random.binomial(n=1, 
                                                            p=grp.p, size=grp.n))
                              }) for grp in obs
                  ],ignore_index=True )

df.to_csv(OUTPUT_LOC,index=False)

Answer 1

老问题，但也许这会对其他人有所帮助。有几个问题。首先是与数据传递给pd.crosstab 的方式有关。 Tableau 将值列表传递给 tabpy 服务器，因此将其包装在一个数组中以修复您遇到的错误。

SCRIPT_REAL(
  "
   import pandas as pd
   import numpy as np
   from scipy.stats import fisher_exact
   df_cont_tbl = pd.crosstab(np.array(_arg1), np.array(_arg2))
   prop_test = fisher_exact(df_cont_tbl)
   return prop_test[1]
   ", 
   attr([Group]), attr([Bookings])

   )

另一个问题是执行表格计算的方式。你想发送 tabpy 两个信息列表，每个列表都和你的表一样长。在默认情况下，tableau 想要在行级别进行计算，这是行不通的。

我将行数 F1 包含在我构建工作簿的 csv 中，并确保沿此函数计算 python 值。

现在，当您将 F1 放入工作表时，它将返回 P 值与您的行数一样多，解决方法是将您的计算包装在另一个计算中，仅在它是第一行时才返回值，并且把它放在你的工作表中。

现在您可以将其放入工作表中。