ó ÎYc@@södZddlmZddlmZddlmZddlmZddlZ ddl Z ddl m Z ddlmZdd lmZdd lmZmZmZmZmZdd lmZd d lmZmZd dlmZmZd dl m!Z!m"Z"m#Z#dgZ$e%e&e j'e j(gZ)e*e j+gZ,ddgZ-dgdgddggZ.defd„ƒYZ/de/fd„ƒYZ0de/fd„ƒYZ1defd„ƒYZ2de3fd„ƒYZ4d„Z5de3fd„ƒYZ6dS( sS The classes and functionality used to transform data inputs to consistent types. i(tabsolute_import(tcopy(tchain(t itemgetterN(t iteritems(tzip(tHasProps(tbokeh_integer_typestDatetimetFloattListtString(tColumnDataSourcei(tColumnt ColumnLabel(tBinstStat(tcollect_attribute_columnstgen_column_namestspecial_columnst _charts_onestxtytColumnAssignercB@sJeZdZeeddƒZeeddƒZdd„Zdd„Z RS(sDefines behavior for assigning columns to dimensions. This class is used to collect assignments between columns and :class:`Builder` dimensions when none are provided. The :class:`ChartDataSource` receives a ColumnAssigner from each :class:`Builder`, which can implement custom behavior. Each subclass must implement the :meth:`get_assignment` method, which returns a `dict` mapping between each dimension in `dims` and one or more column names, or `None` if no assignment is made for the associated dimension. thelpsã The list of dimension names that are associated with the :class:`Builder`. The ColumnAssigner should return a dict with each dimension as a key when the :meth:`get_assignment` method is called. sé This list of attribute names that are associated with the :class:`Builder`. These can be used to alter which dimensions are assigned which columns, versus which attributes are assigned which columns. cK@s2|dk r||_ntt|ƒj|dS(sãCreate the assigner. Args: df (:class:`pandas.DataFrame`, optional): the data source to use for assigning columns from **properties: any attribute of the ColumnAssigner N(tNonet_dftsuperRt__init__(tselftdft properties((s3lib/python2.7/site-packages/bkcharts/data_source.pyR6s  cC@stdƒ‚dS(Ns3You must return map between each dim and selection.(tNotImplementedError(Rt selections((s3lib/python2.7/site-packages/bkcharts/data_source.pytget_assignmentCsN( t__name__t __module__t__doc__R R tdimstattrsRRR"(((s3lib/python2.7/site-packages/bkcharts/data_source.pyRs      tOrderedAssignercB@seZdZdd„ZRS(s’Assigns one column for each dimension that is not an attribute, in order. This is the default column assigner for the :class:`Builder`. cC@s€|dks*tt|jƒƒƒdkrxg|jD]}||jkr4|^q4}d„t||jjj ƒƒDƒS|SdS(sEGet a mapping between dimension and selection when none are provided.icS@si|]\}}||“qS(((t.0tdimtsel((s3lib/python2.7/site-packages/bkcharts/data_source.pys Rs N( RtlentlisttkeysR&R'RRtcolumnsttolist(RR!R*R&((s3lib/python2.7/site-packages/bkcharts/data_source.pyR"Ns *+N(R#R$R%RR"(((s3lib/python2.7/site-packages/bkcharts/data_source.pyR(GstNumericalColumnsAssignercB@seZdZdd„ZRS(s1Assigns all numerical columns to the y dimension.cC@sèt|tƒr0|jdƒ}|jdƒ}nd}d}i}|jjƒ}|jjƒ}|dk r¦|dkr¦g|D]}|t|ƒkr|^q}n*|dkrÐd}|dkrÐ|}qÐn||d<||d<|S(NRRtindex( t isinstancetdicttgetRRt_get_numeric_dataR/R0R-(RR!RRRtnum_colstcol((s3lib/python2.7/site-packages/bkcharts/data_source.pyR"[s".     N(R#R$R%RR"(((s3lib/python2.7/site-packages/bkcharts/data_source.pyR1Xst DataOperatorcB@s;eZdZeeƒddddƒZd„Zd„ZRS(sAAn operation that transforms data before it is used for plotting.tdefaultRs1 List of columns to perform operation on.cC@stdƒ‚dS(Ns3Each data operator must implement the apply method.(R (Rtdata((s3lib/python2.7/site-packages/bkcharts/data_source.pytapplyzscC@s&dj|jƒ}d|jj|fS(Ns, s%s(%s)(tjoinR/t __class__R#(Rtcol_str((s3lib/python2.7/site-packages/bkcharts/data_source.pyt__repr__}sN( R#R$R%R RRR/R<R@(((s3lib/python2.7/site-packages/bkcharts/data_source.pyR9us   t DataGroupcB@sbeZdZd„Zd„Zed„ƒZd„Zd„Zd„Z ed„ƒZ d„Z RS( sÝContains subset of data and metadata about it. The DataGroup contains a map from the labels of each attribute associated with an :class:`AttrSpec` to the value of the attribute assigned to the DataGroup. cC@s||_||_||_dS(sCreate a DataGroup for the data, with a label and associated attributes. Args: label (str): the label for the group based on unique values of each column data (:class:`pandas.DataFrame`): the subset of data associated with the group attr_specs dict(str, :class:`AttrSpec`): mapping between attribute name and the associated :class:`AttrSpec`. N(tlabelR;t attr_specs(RRBR;RC((s3lib/python2.7/site-packages/bkcharts/data_source.pyRŠs  cC@s~t|tƒr|j|St|tƒrJt|ƒdkrJ|j|dSt|tƒrvt|ƒdkrv|j|SdSdS(sÐGet the data associated with the selection of columns. Args: selection (List(Str) or Str): the column or columns selected Returns: :class:`pandas.DataFrame` iiN(R3tstrR;R-R,R(Rt selection((s3lib/python2.7/site-packages/bkcharts/data_source.pyt get_values˜s  !! cC@s t|jƒS(s>The :class:`ColumnDataSource` representation of the DataFrame.(R R;(R((s3lib/python2.7/site-packages/bkcharts/data_source.pytsource«scC@s |j|S(sCGet the value of the :class:`AttrSpec` associated with `spec_name`.(RC(Rt spec_name((s3lib/python2.7/site-packages/bkcharts/data_source.pyt __getitem__°scC@sdt|jƒ|jfS(Ns (RDRBRC(R((s3lib/python2.7/site-packages/bkcharts/data_source.pyR@´scC@st|jjƒS(N(R,R;R2(R((s3lib/python2.7/site-packages/bkcharts/data_source.pyt__len__·scC@st|jjƒƒS(N(R-RCR.(R((s3lib/python2.7/site-packages/bkcharts/data_source.pyt attributesºscC@s_i}|jdk rA|j|jƒt|jjƒƒ|dIs (RRj(R((Rs3lib/python2.7/site-packages/bkcharts/data_source.pyRCGscK@si}x<|jD]1}|j|dƒ}|dk r|||Rftappend(RR!R*tselect((s3lib/python2.7/site-packages/bkcharts/data_source.pyRluscC@s~xw|jD]l}||}|dk r t|tƒr |tkr ||jjjƒkr t||jƒ|j|t t |j jƒt |ƒƒ}t j |j d|d|d|d|ƒ|_ q q WdS(sÕ De-pivots `_data` from a 'wide' to 'tall' layout. A wide table is one where the column names represent a categorical variable and each contains only the values associated with each unique value of the categorical variable. This method uses the :func:`pandas.melt` function with additional logic to make sure that the same data source can have multiple operations applied, and so all other columns are maintained through the stacking process. Example: .. note:: This example is fairly low level and is not something the typical user should worry about. The interface for data transformations from the user perspective are the :ref:`bkcharts_functions`. >>> data = {'a': [1, 2, 3, 4], ... 'b': [2, 3, 4, 5], ... 'month': ['jan', 'jan', 'feb', 'feb'] ... } >>> ds = ChartDataSource.from_data(data) >>> ds['x'] =['a', 'b'] # say we selected a and b for dimension x We may want to combine 'a' and 'b' together. The final data would look like the following: >>> ds.stack_measures(['c', 'd'], var_name='c_d_variable', ... value_name='c_d_value') >>> ds.df Out[35]: month a_b_variable a_b_value 0 jan a 1 1 jan a 2 2 feb a 3 3 feb a 4 4 jan b 2 5 jan b 3 6 feb b 4 7 feb b 5 The transformed data will use the `var_name` and `value_name` inputs to name the columns. These derived columns can then be used as a single column to reference the values and the labels of the data. In the example, I could plot a_b_value vs month, and color by a_b_variable. What this does for you over the :meth:`pandas.melt` method is that it will apply the :class:`DataOperator` for a dimension if it exists (e.g. :class:`Blend`, generated by :func:`blend`), and it will try to handle the id columns for you so you don't lose other columns with the melt transformation. Returns: None t_tid_varst value_varstvar_namet value_nameN(ReRjR3R9R/RR0RtalltpdtmeltRdR-tset( RtmeasurestidsR€RR*REtdim_colstmeasure((s3lib/python2.7/site-packages/bkcharts/data_source.pytstack_measures©s <      1"cK@s1t|ƒdkr!tdƒ‚nt|j|S(sN Iterable of chart attribute specifications, associated with columns. Iterates over DataGroup, which represent the lowest level of data that is assigned to the attributes for plotting. Yields: a DataGroup, which contains metadata and attributes assigned to the group of data isBYou must provide one or more Attribute Specs to support iteration.(R,RsRTRd(RRV((s3lib/python2.7/site-packages/bkcharts/data_source.pyRTs  c K@s-|jjƒ}|s!|j}ng}g}t}xO|j|D]>}|jdkrat}n|j|ƒ|j|j ƒƒqCW|ràt j j |dj ƒgƒ}d|dósR&icS@si|]\}}||“qS(((R)t column_nameR³((s3lib/python2.7/site-packages/bkcharts/data_source.pys s RR;N(tanyRR-RR¡RR,R5RaRRct enumerateR3RƒR¢RXRR‘R¥(R§Rªt column_namesRqtlist_of_arraysR\R&R*t list_of_arrayR}R+tiR³RXRµ((R§s3lib/python2.7/site-packages/bkcharts/data_source.pyR£ês"   " cK@s|dtjj|ƒ|S(syProduce :class:`ColumnDataSource` from table-like dict. Returns: :class:`ColumnDataSource` R(RƒR‘R¥(R§R;Rq((s3lib/python2.7/site-packages/bkcharts/data_source.pyR¥ scC@stj|tƒptj|ƒS(s}Verify if data is table-like. Inspects the types and structure of data. Returns: bool (R^t _is_validt TABLE_TYPESRŸ(R;((s3lib/python2.7/site-packages/bkcharts/data_source.pyRžs cC@s5t|tƒo4tg|D]}t|tƒ^qƒS(s]Verify if data is row-oriented, table-like data. Returns: bool (R3R-R‚R4(R;RP((s3lib/python2.7/site-packages/bkcharts/data_source.pyRŸ"scC@s'tj|ƒrtStj|tƒSdS(sJVerify if data is array-like. Returns: bool N(R^RŸRURÃt ARRAY_TYPES(R;((s3lib/python2.7/site-packages/bkcharts/data_source.pyR,scC@s&tg|D]}t||ƒ^q ƒS(s®Checks for each type against data. Args: data: a generic source of data types: a list of classes Returns: bool (R½R3(R;ttypest valid_type((s3lib/python2.7/site-packages/bkcharts/data_source.pyRÃ:s c C@s·|j}|j}gt|ƒD]\}}|dk r|^q}|jdk r}g|D]}||jkrY|^qY}nt|ƒdkr¯xU|D]M}t|ƒt|ƒkrºq–ntg|D]}||k^qăr–dSq–Wddd}g|D],}djg|D]}d|^qƒ^qü} gt|ƒD]4\}}|dk r;dt|ƒt|ƒf^q;} t |d j| ƒd j| ƒd j|j ƒfƒ‚ndSdS( syRaises selection error if selections are not valid compared to requirements. Returns: None iNsPDid not receive a valid combination of selections. Valid configurations are: %ss Received inputs are: %ss Available columns are: %ss and s%s = s%s = %ss or s, ( RgRjRRR'R,R‚R=RDRsR/( RRpR!R*R+R&treqt error_strt required_dimtreq_strt selection_str((s3lib/python2.7/site-packages/bkcharts/data_source.pyRoHs(  1+ %6D4cC@sCt|tjƒr%ttƒj|ƒStft}t||ƒSdS(sUVerifies that value is a numerical type. Returns: bool N(R3RƒR¢R R tis_validtfloatR(R]tnumbers((s3lib/python2.7/site-packages/bkcharts/data_source.pyt is_numberss cC@s1yt|ƒ}|tSWntk r,tSXdS(suVerifies that value is a valid Datetime type, or can be converted to it. Returns: bool N(RRRsRU(R]tdt((s3lib/python2.7/site-packages/bkcharts/data_source.pyt is_datetimes   cC@siS(s6Introspect which columns match to which types of data.((R;((s3lib/python2.7/site-packages/bkcharts/data_source.pyRm‘scC@s |jjS(s\All column names associated with the data. Returns: List(Str) (RdR/(R((s3lib/python2.7/site-packages/bkcharts/data_source.pyR/—scC@s |jjS(s8The index for the :class:`pandas.DataFrame` data source.(RdR2(R((s3lib/python2.7/site-packages/bkcharts/data_source.pyR2¡scC@s |jjS(N(RdR¤(R((s3lib/python2.7/site-packages/bkcharts/data_source.pyR¤¦scC@s|tkrtStSdS(smVerify if the column provided matches to known computed columns. Returns: bool N(tCOMPUTED_COLUMN_NAMESRRU(tcolumn((s3lib/python2.7/site-packages/bkcharts/data_source.pyt is_computedªs N(%R#R$R%RR(RRRRCRiRlRkRIR{RŠRTR™t classmethodR¶t staticmethodR RRGR»R£R¥RžRŸRRÃRoRÐRÒRmR/R2R¤RÕ(((s3lib/python2.7/site-packages/bkcharts/data_source.pyR^ sB  * *    X  -x     + (7R%t __future__RRt itertoolsRtoperatorRtnumpyR·tpandasRƒtsixRt six.movesRtbokeh.core.has_propsRtbokeh.core.propertiesRRR R R tbokeh.models.sourcesR RR RtstatsRRtutilsRRRRÓRNR-R¸R¢RÅR4R‘RÄRaRbRR(R1R9tobjectRARTR^(((s3lib/python2.7/site-packages/bkcharts/data_source.pyts4  (  ( H C