VariablesToTable Class Reference

Inheritance diagram for VariablesToTable:

Public Member Functions
def	__init__ (self, str listname, List[str] variables, str filename, Optional[str] hdf_table_name=None, int event_buffer_size=100, **writer_kwargs)
def	initialize (self)
def	buffer (self)
def	event_buffer (self)
def	clear_buffer (self)
def	append_buffer (self)
def	initialize_feather_writer (self)
def	initialize_parquet_writer (self)
def	initialize_csv_writer (self)
def	initialize_hdf5_writer (self)
def	fill_event_buffer (self)
def	buffer_full (self)
def	write_buffer (self)
def	event (self)
def	terminate (self)

Protected Attributes
	_filename
	Output filename.
	_listname
	Particle list name.
	_variables
	List of variables.
	_format
	Output format.
	_table_name
	Table name in the hdf5 file.
	_event_buffer_size
	Event buffer size.
	_event_buffer_counter
	Event buffer counter.
	_writer_kwargs
	writer kwargs
	_varnames
	variable names
	_std_varnames
	std::vector of variable names
	_evtmeta
	Event metadata.
	_plist
	Pointer to the particle list.
	_dtypes
	The data type.
	_buffer
	event variables buffer (will be automatically grown if necessary)
	_buffer_index
	current start index in the event variables buffer
	_schema
	A list of tuples and py.DataTypes to define the pyarrow schema.
	_feather_writer
	a writer object to write data into a feather file
	_parquet_writer
	a writer object to write data into a parquet file
	_csv_writer
	a writer object to write data into a csv file
	_hdf5_writer
	The pytable file.
	_table
	The pytable.

Detailed Description

Base class to dump ntuples into a non root format of your choosing

Definition at line 38 of file b2pandas_utils.py.

Constructor & Destructor Documentation

__init__()

def __init__	(		self,
		str	listname,
		List[str]	variables,
		str	filename,
		Optional[str]	hdf_table_name = None,
		int	event_buffer_size = 100,
		**	writer_kwargs
	)

Constructor to initialize the internal state

Arguments:
    listname(str): name of the particle list
    variables(list(str)): list of variables to save for each particle
    filename(str): name of the output file to be created.
        Needs to end with `.csv` for csv output, `.parquet` or `.pq` for parquet output,
        `.h5`, `.hdf` or `.hdf5` for hdf5 output and `.feather` or `.arrow` for feather output
    hdf_table_name(str): name of the table in the hdf5 file.
        If not provided, it will be the same as the listname
    event_buffer_size(int): number of events to buffer before writing to disk,
        higher values will use more memory but write faster and result in smaller files
    **writer_kwargs: additional keyword arguments to pass to the writer.
        For details, see the documentation of the writer in the apache arrow documentation.
        Only use, if you know what you are doing!

Reimplemented in VariablesToHDF5.

Definition at line 43 of file b2pandas_utils.py.

    ):
        """Constructor to initialize the internal state
 
        Arguments:
            listname(str): name of the particle list
            variables(list(str)): list of variables to save for each particle
            filename(str): name of the output file to be created.
                Needs to end with `.csv` for csv output, `.parquet` or `.pq` for parquet output,
                `.h5`, `.hdf` or `.hdf5` for hdf5 output and `.feather` or `.arrow` for feather output
            hdf_table_name(str): name of the table in the hdf5 file.
                If not provided, it will be the same as the listname
            event_buffer_size(int): number of events to buffer before writing to disk,
                higher values will use more memory but write faster and result in smaller files
            **writer_kwargs: additional keyword arguments to pass to the writer.
                For details, see the documentation of the writer in the apache arrow documentation.
                Only use, if you know what you are doing!
        """
        super().__init__()
        
        self._filename = filename
        
        self._listname = listname
        
        self._variables = list(set(variables))
        
        file_type = self._filename.split(".")[-1]
        if file_type in ["csv"]:
            self._format = "csv"
        elif file_type in ["parquet", "pq"]:
            self._format = "parquet"
        elif file_type in ["h5", "hdf", "hdf5"]:
            self._format = "hdf5"
        elif file_type in ["feather", "arrow"]:
            self._format = "feather"
        else:
            raise ValueError(
                f"Unknown file type ending .{file_type}, supported types are 'csv', "
                "'parquet', 'pq', 'h5', 'hdf', 'hdf5', 'feather' or 'arrow'"
            )
        
        self._table_name = (
            hdf_table_name if hdf_table_name is not None else self._listname
        )
        
        self._event_buffer_size = event_buffer_size
        
        self._event_buffer_counter = 0
        
        self._writer_kwargs = writer_kwargs
 

Member Function Documentation

append_buffer()

def append_buffer

(

self

)

"Append" a new event to the buffer by moving the buffer index forward by particle list size

Automatically replaces the buffer by a larger one if necessary

Definition at line 178 of file b2pandas_utils.py.

    def append_buffer(self):
        """
        "Append" a new event to the buffer by moving the buffer index forward by particle list size
 
        Automatically replaces the buffer by a larger one if necessary
        """
        plist_size = self._plist.getListSize()
        if (plist_size + self._buffer_index) > len(self._buffer):
            new_buffer = np.empty(
                # factor 1.5 larger or at least as large as necessary
                max(int(len(self._buffer) * 1.5), self._buffer_index + plist_size),
                dtype=self._dtypes,
            )
            new_buffer[:self._buffer_index] = self.buffer
            self._buffer = new_buffer
        self._buffer_index += plist_size
        self._event_buffer_counter += 1
 

buffer()

def buffer

(

self

)

The buffer slice across multiple entries

Definition at line 158 of file b2pandas_utils.py.

    def buffer(self):
        """
        The buffer slice across multiple entries
        """
        return self._buffer[:self._buffer_index]
 

buffer_full()

def buffer_full

(

self

)

check if the buffer is full

Definition at line 275 of file b2pandas_utils.py.

    def buffer_full(self):
        """
        check if the buffer is full
        """
        return self._event_buffer_counter == self._event_buffer_size
 

clear_buffer()

def clear_buffer

(

self

)

Reset the buffer event counter and index

Definition at line 171 of file b2pandas_utils.py.

    def clear_buffer(self):
        """
        Reset the buffer event counter and index
        """
        self._event_buffer_counter = 0
        self._buffer_index = 0
 

event()

def event

(

self

)

Event processing function

executes the fill_buffer function and writes the data to the output file
in chunks of event_buffer_size

Definition at line 298 of file b2pandas_utils.py.

    def event(self):
        """
        Event processing function
 
        executes the fill_buffer function and writes the data to the output file
        in chunks of event_buffer_size
        """
        self.append_buffer()
        self.fill_event_buffer()
        if self.buffer_full:
            self.write_buffer()
            self.clear_buffer()
 

event_buffer()

def event_buffer

(

self

)

The buffer slice for the current event

Definition at line 165 of file b2pandas_utils.py.

    def event_buffer(self):
        """
        The buffer slice for the current event
        """
        return self._buffer[self._buffer_index - self._plist.getListSize(): self._buffer_index]
 

fill_event_buffer()

def fill_event_buffer

(

self

)

Assign values for all variables for all particles in the particle list to the current event buffer

Definition at line 254 of file b2pandas_utils.py.

    def fill_event_buffer(self):
        """
        Assign values for all variables for all particles in the particle list to the current event buffer
        """
        buf = self.event_buffer
 
        # add some extra columns for bookkeeping
        buf["__experiment__"] = self._evtmeta.getExperiment()
        buf["__run__"] = self._evtmeta.getRun()
        buf["__event__"] = self._evtmeta.getEvent()
        buf["__production__"] = self._evtmeta.getProduction()
        buf["__ncandidates__"] = len(buf)
        buf["__candidate__"] = np.arange(len(buf))
 
        # fill variables into buffer
        vector = variables.variables.evaluateVariables(self._std_varnames, self._plist)
        values = np.array(vector.data()).reshape(-1, len(self._varnames))
        for name, col in zip(self._varnames, values.T):
            buf[name] = col
 

initialize()

def initialize

(

self

)

Setup variable lists, pointers, buffers and file writers

Definition at line 101 of file b2pandas_utils.py.

    def initialize(self):
        """
        Setup variable lists, pointers, buffers and file writers
        """
        # Always avoid the top-level 'import ROOT'.
        import ROOT  # noqa
 
        
        self._varnames = [
            str(varname)
            for varname in variables.variables.resolveCollections(
                variables.std_vector(*self._variables)
            )
        ]
 
        
        self._std_varnames = variables.std_vector(*self._varnames)
 
        
        self._evtmeta = ROOT.Belle2.PyStoreObj("EventMetaData")
        self._evtmeta.isRequired()
 
        
        self._plist = ROOT.Belle2.PyStoreObj(self._listname)
        self._plist.isRequired()
 
        dtypes = [
            ("__experiment__", np.int32),
            ("__run__", np.int32),
            ("__event__", np.uint32),
            ("__production__", np.uint32),
            ("__candidate__", np.uint32),
            ("__ncandidates__", np.uint32),
        ]
        for name in self._varnames:
            # only float variables for now
            dtypes.append((name, np.float64))
 
        
        self._dtypes = dtypes
 
        
        self._buffer = np.empty(self._event_buffer_size * 10, dtype=self._dtypes)
 
        
        self._buffer_index = 0
 
        if self._format == "hdf5":
            self.initialize_hdf5_writer()
        elif self._format == "parquet":
            self.initialize_parquet_writer()
        elif self._format == "csv":
            self.initialize_csv_writer()
        elif self._format == "feather":
            self.initialize_feather_writer()
 

initialize_csv_writer()

def initialize_csv_writer

(

self

)

Initialize the csv writer using pyarrow

Definition at line 224 of file b2pandas_utils.py.

    def initialize_csv_writer(self):
        """
        Initialize the csv writer using pyarrow
        """
        
        self._schema = [
            (name, numpy_to_pyarrow_type_map[dt]) for name, dt in self._dtypes
        ]
        
        self._csv_writer = CSVWriter(self._filename, schema=pa.schema(self._schema), **self._writer_kwargs)
 

initialize_feather_writer()

def initialize_feather_writer

(

self

)

Initialize the feather writer using pyarrow

Definition at line 196 of file b2pandas_utils.py.

    def initialize_feather_writer(self):
        """
        Initialize the feather writer using pyarrow
        """
        
        self._schema = [
            (name, numpy_to_pyarrow_type_map[dt]) for name, dt in self._dtypes
        ]
        
        self._feather_writer = ipc.RecordBatchFileWriter(
            sink=self._filename,
            schema=pa.schema(self._schema),
            **self._writer_kwargs,
        )
 

initialize_hdf5_writer()

def initialize_hdf5_writer

(

self

)

Initialize the hdf5 writer using pytables

Definition at line 235 of file b2pandas_utils.py.

    def initialize_hdf5_writer(self):
        """
        Initialize the hdf5 writer using pytables
        """
        
        self._hdf5_writer = tables.open_file(
            self._filename, mode="w", title="Belle2 Variables to HDF5"
        )
        filters = tables.Filters(complevel=1, complib="blosc:lz4", fletcher32=False)
 
        # some variable names are not just A-Za-z0-9 so pytables complains but
        # seems to work. Ignore warning
        with warnings.catch_warnings():
            warnings.simplefilter("ignore")
            
            self._table = self._hdf5_writer.create_table(
                "/", self._table_name, obj=np.zeros(0, self._dtypes), filters=filters, **self._writer_kwargs
            )
 

initialize_parquet_writer()

def initialize_parquet_writer

(

self

)

Initialize the parquet writer using pyarrow

Definition at line 211 of file b2pandas_utils.py.

    def initialize_parquet_writer(self):
        """
        Initialize the parquet writer using pyarrow
        """
        
        self._schema = [
            (name, numpy_to_pyarrow_type_map[dt]) for name, dt in self._dtypes
        ]
        
        self._parquet_writer = ParquetWriter(
            self._filename, schema=pa.schema(self._schema), **self._writer_kwargs
        )
 

terminate()

def terminate

(

self

)

save and close the output

Definition at line 311 of file b2pandas_utils.py.

    def terminate(self):
        """save and close the output"""
        import ROOT  # noqa
        if len(self.buffer) > 0:
            self.write_buffer()
 
        if self._format == "hdf5":
            self._table.flush()
            self._hdf5_writer.close()
        elif self._format == "parquet":
            self._parquet_writer.close()
        elif self._format == "csv":
            self._csv_writer.close()
        elif self._format == "feather":
            self._feather_writer.close()
        ROOT.Belle2.MetadataService.Instance().addNtuple(self._filename)
 
 

write_buffer()

def write_buffer

(

self

)

write the buffer to the output file

Definition at line 281 of file b2pandas_utils.py.

    def write_buffer(self):
        """
        write the buffer to the output file
        """
        if self._format == "hdf5":
            """Create a new row in the hdf5 file with for each particle in the list"""
            self._table.append(self.buffer)
        else:
            table = {name: self.buffer[name] for name, _ in self._dtypes}
            pa_table = pa.table(table, schema=pa.schema(self._schema))
            if self._format == "parquet":
                self._parquet_writer.write_table(pa_table)
            elif self._format == "csv":
                self._csv_writer.write(pa_table)
            elif self._format == "feather":
                self._feather_writer.write_table(pa_table)
 

Member Data Documentation

_buffer

_buffer

protected

event variables buffer (will be automatically grown if necessary)

Definition at line 143 of file b2pandas_utils.py.

_buffer_index

_buffer_index

protected

current start index in the event variables buffer

Definition at line 146 of file b2pandas_utils.py.

_csv_writer

_csv_writer

protected

a writer object to write data into a csv file

Definition at line 233 of file b2pandas_utils.py.

_dtypes

_dtypes

protected

The data type.

Definition at line 140 of file b2pandas_utils.py.

_event_buffer_counter

_event_buffer_counter

protected

Event buffer counter.

Definition at line 97 of file b2pandas_utils.py.

_event_buffer_size

_event_buffer_size

protected

Event buffer size.

Definition at line 95 of file b2pandas_utils.py.

_evtmeta

_evtmeta

protected

Event metadata.

Definition at line 120 of file b2pandas_utils.py.

_feather_writer

_feather_writer

protected

a writer object to write data into a feather file

Definition at line 205 of file b2pandas_utils.py.

_filename

_filename

protected

Output filename.

Definition at line 70 of file b2pandas_utils.py.

_format

_format

protected

Output format.

Definition at line 78 of file b2pandas_utils.py.

_hdf5_writer

_hdf5_writer

protected

The pytable file.

Definition at line 240 of file b2pandas_utils.py.

_listname

_listname

protected

Particle list name.

Definition at line 72 of file b2pandas_utils.py.

_parquet_writer

_parquet_writer

protected

a writer object to write data into a parquet file

Definition at line 220 of file b2pandas_utils.py.

_plist

_plist

protected

Pointer to the particle list.

Definition at line 124 of file b2pandas_utils.py.

_schema

_schema

protected

A list of tuples and py.DataTypes to define the pyarrow schema.

Definition at line 201 of file b2pandas_utils.py.

_std_varnames

_std_varnames

protected

std::vector of variable names

Definition at line 117 of file b2pandas_utils.py.

_table

_table

protected

The pytable.

Definition at line 250 of file b2pandas_utils.py.

_table_name

_table_name

protected

Table name in the hdf5 file.

Definition at line 91 of file b2pandas_utils.py.

_variables

_variables

protected

List of variables.

Definition at line 74 of file b2pandas_utils.py.

_varnames

_varnames

protected

variable names

Definition at line 109 of file b2pandas_utils.py.

_writer_kwargs

_writer_kwargs

protected

writer kwargs

Definition at line 99 of file b2pandas_utils.py.

---

The documentation for this class was generated from the following file:

• analysis/scripts/b2pandas_utils.py