我有一个用pytables创建的数据集,我试图将其导入到pandas数据帧中。我无法将where
过滤器应用于read_hdf
步骤。我在熊猫'0.12.0'
我的样本pytables数据:
import tables
import pandas as pd
import numpy as np
class BranchFlow(tables.IsDescription):
branch = tables.StringCol(itemsize=25, dflt=' ')
flow = tables.Float32Col(dflt=0)
filters = tables.Filters(complevel=8)
h5 = tables.openFile('foo.h5', 'w')
tbl = h5.createTable('/', 'BranchFlows', BranchFlow,
'Branch Flows', filters=filters, expectedrows=50e6)
for i in range(25):
element = tbl.row
element['branch'] = str(i)
element['flow'] = np.random.randn()
element.append()
tbl.flush()
h5.close()
我可以将其导入数据帧:
store = pd.HDFStore('foo.h5')
print store
print pd.read_hdf('foo.h5', 'BranchFlows').head()
显示:
In [10]: print store
<class 'pandas.io.pytables.HDFStore'>
File path: foo.h5
/BranchFlows frame_table [0.0.0] (typ->generic,nrows->25,ncols->2,indexers->[index],dc->[branch,flow])
In [11]: print pd.read_hdf('foo.h5', 'BranchFlows').head()
branch flow
0 0 -0.928300
1 1 -0.256454
2 2 -0.945901
3 3 1.090994
4 4 0.350750
但我无法让过滤器在流列上工作:
pd.read_hdf('foo.h5', 'BranchFlows', where=['flow>0.5'])
<snip traceback>
TypeError: passing a filterable condition to a non-table indexer [field->flow,op->>,value->[0.5]]
答案 0 :(得分:3)
从PyTables直接创建的表中读取只允许您直接读取(整个)表。您必须使用pandas工具(表格式)编写它以使用pandas选择机制(因为pandas需要的元数据不存在 - 可以完成,但需要一些工作)。
因此,请像上面一样阅读您的表格,然后创建一个新表格,并指明表格格式。见here for docs
In [6]: df.to_hdf('foo.h5','BranchFlowsTable',data_columns=True,table=True)
In [24]: with pd.get_store('foo.h5') as store:
print(store)
....:
<class 'pandas.io.pytables.HDFStore'>
File path: foo.h5
/BranchFlows frame_table [0.0.0] (typ->generic,nrows->25,ncols->2,indexers->[index],dc->[branch,flow])
/BranchFlowsTable frame_table (typ->appendable,nrows->25,ncols->2,indexers->[index],dc->[branch,flow])
In [7]: pd.read_hdf('foo.h5','BranchFlowsTable',where='flow>0.5')
Out[7]:
branch flow
14 14 1.503739
15 15 0.660297
17 17 0.685152
18 18 1.156073
20 20 0.994792
21 21 1.266463
23 23 0.927678