datatable_alt.py

'''
alternate datatable implementation using dict of lists
provides a proxy dict implementation which provides read+write-through access to the data by row
'''
from collections import defaultdict
from datatable_util import AttributeDict, DataTableException, CSV_GivenHeaders, FIXEDWIDTH, JoinType, sortKey
from hierarchies import Hierarchy
import os
from functools import total_ordering

@total_ordering
class DataColumn(object):
	def __init__(self, dataTable, header, data=[]):
		self.__dataTable = dataTable
		if isinstance(header, DataColumn):
			self.header = header.header
			self.__data = list(header.__data)
		else:
			self.header = header
			self.__data = []
			for item in data:
				if isinstance(item, (dict, DataRowProxy)):
					if header in item:
						self.__data.append(item[header])
					else:
						self.__data.append(None)
				else:
					self.__data.append(item)
	def __eq__(self, other):
		if not isinstance(other, DataColumn):
			return False
		return self.__data == other.__data
	def __lt__(self, other):
		if not isinstance(other, DataColumn) or self.__dataTable is not other.__dataTable:
			raise NotImplemented()
		return sortKey(self.header) < sortKey(other.header)
	def __iter__(self):
		return iter(self.__data)
	def __len__(self):
		return len(self.__data)
	def __getitem__(self, index):
		'''Gets the index'th row of data'''
		return self.__data[index]
	def __contains__(self, value):
		return value in self.__data
	def __filter(self, value):
		if value is None:
			for i in range(len(self)):
				if self[i] is None:
					yield i
		elif isinstance(value, DataColumn):
			if value.__dataTable == self.__dataTable:
				for i in range(len(self)):
					if self[i] == value[i]:
						yield i
			else:
				otherValues = set(value)
				for i in range(len(self)):
					if self[i] in otherValues:
						yield i
		elif '__call__' in dir(value):
			for i in range(len(self)):
				if value(self[i]):
					yield i
		elif '__contains__' in dir(value) and not isinstance(value, str):
			for i in range(len(self)):
				if self[i] in value:
					yield i
		else:
			for i in range(len(self)):
				if self[i] == value:
					yield i
	def filter(self, value):
		'''
	Filter the table by matching this column with the given value
Value may be one of the following:
	None - returns rows where this column is None
	DataColumn (same table) - returns rows where the two columns are equal
	DataColumn (other table) - returns rows where this column value is in the other column
	method - returns rows where method returns true for column value
	collection - returns rows where column value is in the collection
	value - returns rows where column value equals the given value
		'''
		return self.__dataTable.select(self.__filter(value))
	def set(self, value):
		'''
	sets the items in this column to the given value
	If value is a function then sets each item to the result of calling value on the item
	returns the modified datatable
'''
		if hasattr(value, '__call__'):
			self.__data = [value(data) for data in self]
		else:
			self.__data = [value] * len(self)
		return self.__dataTable
	def sort(self):
		self.__dataTable.sort(self.header)
	def sizeOfGroups(self):
		groups = defaultdict(lambda:0)
		for v in self:
			groups[v] += 1
		return dict(groups)
	def fillDownBlanks(self):
		prev = None
		for i in range(len(self.__data)):
			if self.__data[i]:
				prev = self.__data[i]
			elif prev:
				self.__data[i] = prev
	def __repr__(self):
		return "DataColumn(<dataTable>, '%s')" % self.header
	def __str__(self):
		return ','.join(map(str, self))

class NullColumn(DataColumn):
	def __filter(self, value):
		return iter([])
	def sort(self):
		pass
	def sizeOfGroups(self):
		return {}
	def __iter__(self):
		return iter([])
	def __repr__(self):
		return "NullColumn(<dataTable>, '%s')" % self.header

def _copyAndApplyOp(op):
	def copyOp(self, *args, **kwargs):
		newData = DataTable(self)
		op(newData, *args, **kwargs)
		return newData
	copyOp.__doc__ = op.__doc__
	return copyOp

class DataTable(object):
	@staticmethod
	def collect(tables):
		'''
	Concatenates the tables together into one big data table
	essentially performs:
table = tables.next()
for t in tables:
	table.augment(t)
		'''
		data = {}
		size = 0
		for table in tables:
			for column in table.columns():
				if column.header not in data:
					data[column.header] = [None]*size
				data[column.header] += list(column)
			for h in data.keys():
				if h not in table.headers():
					data[h] += [None] * len(table)
			size += len(table)
		return DataTable(DataColumn(None, header, columnData) for header, columnData in data.items())
	def __init__(self, data=None, parseMethod=None):
		'''Create a data table from the given data
	data may be one of the following:
A sequence of dictionaries, where all of the dictionaries share common keys
A sequence of sequences where the first item is the list of headers
Another DataTable instance, which will create a deep copy
A string which may be parsed into one of the previous by calling parseMethod on the string.
'''
		if isinstance(data, DataTable):
			self.__headers = {h: DataColumn(self, c) for h, c in data.__headers.items()}
			self.__length = len(data)
			return
		if isinstance(data, str):
			data = parseMethod(data)
		if not data:
			self.__headers = {}
			self.__length = 0
			return
		data = [row for row in data]
		if not data:
			self.__headers = {}
			self.__length = 0
			return
		if isinstance(data[0], DataColumn):
			self.__headers = {column.header: DataColumn(self, column) for column in data}
			self.__length = len(data[0])
			return
		if isinstance(data[0], (dict, DataRowProxy)):
			headers = {k for row in data for k in row.keys()}
			self.__headers = {h: DataColumn(self, h, data) for h in headers}
		else:
			headers = data.pop(0)
			data = [{headers[i]: row[i] for i in range(len(headers))} for row in data]
			self.__headers = {h: DataColumn(self, h, data) for h in headers}
		self.__length = len(data)
	def getRow(self, index):
		return DataRowProxy(self, index)
	def select(self, indices):
		if isinstance(indices, slice):
			start, stop, step = indices.indices(len(self)) #calculate the actual start, stop, step given the current length
			indices = range(start, stop, step)
		else:
			indices = list(indices)
		return DataTable(DataColumn(None, c.header, (c[i] for i in indices)) for c in self.__headers.values())
	def __iter__(self):
		'''Gets an iterator over the data rows'''
		for i in range(len(self)):
			yield self.getRow(i)
	def __getitem__(self, index):
		'''Gets the index'th row of data'''
		if isinstance(index, int):
			return self.getRow(index)
		return self.select(index)
	def column(self, header):
		'''Gets the column named 'header' (same as dataTable.<header>)'''
		if header in self.__headers:
			return self.__headers[header]
		return NullColumn(self, header)
	def __getattr__(self, header):
		return self.column(header)
	def columns(self):
		'''Returns the DataColumn objects associated with this DataTable'''
		return sorted(self.__headers.values())
	def headers(self):
		'''Returns this table's header strings'''
		return sorted(self.__headers.keys(), key=sortKey)
	def filter(self, filterFunction):
		'''Returns a DataTable containing the lines in self filtered by the given filterFunciton
	Accepts either a dictionary of header -> value which does exact matching on the pairs,
	or a filter function which takes a dict as input and returns if that row should be included'''
		if isinstance(filterFunction, dict):
			return self.select(i for i in range(len(self)) if all(self.column(k)[i] == v for k, v in filterFunction.items()))
		return DataTable(line for line in self if filterFunction(line))
	def __len__(self):
		'''The number of rows'''
		return self.__length
	def index(self, keyHeaders, leafHeaders=None):
		if leafHeaders is None:
			leafHeaders = set(self.headers()).difference(keyHeaders)
		return Hierarchy.fromTable(self, keyHeaders, leafHeaders)
	def __str__(self):
		return self | FIXEDWIDTH
	def __repr__(self):
		return 'Rows:%d\nHeaders:\n%s' % (len(self), self.headers())
	def __eq__(self, other):
		return isinstance(other, DataTable) and len(self) == len(other) and all(a == b for a, b in zip(self, other))
	def __ne__(self, other):
		return not isinstance(other, DataTable) or len(self) != len(other) or any(a != b for a,b in zip(self, other))
	def augment(self, other):
		'''Join two DataTable instances (concatenate their rows)
	if the headers don't match between the two instances then it adds blank columns to each with the headers from the other'''
		if not other or not len(other):
			return self
		if isinstance(other, list):
			other = DataTable(other)
		if isinstance(other, dict):
			other = DataTable([other])
		if not len(self):
			return other
		selfNewHeaders = {h: None for h in other.headers() if h not in self.headers()}
		otherNewHeaders = {h: None for h in self.headers() if h not in other.headers()}
		return (self & selfNewHeaders) + (other & otherNewHeaders)
	def __iadd__(self, other):
		'''Join two DataTable instances (concatenate their rows)
	requires that the headers match (or that one of self or other be empty)'''
		if other is None:
			return self
		if isinstance(other, DataTable):
			if self.headers() and other.headers() and self.headers() != other.headers():
				raise DataTableException("headers don't match.  Expected: " + str(self.headers()) + "\nFound: " + str(other.headers()))
			for h in self.headers():
				self.__headers[h] = DataColumn(self, h, list(self.__headers[h]) + list(other.__headers[h]))
			self.__length = self.__length + len(other)
		elif isinstance(other, list):
			if other and self.headers() != sorted(other[0].keys(), key=sortKey):
				raise DataTableException("headers don't match.  Expected: " + str(self.headers()) + "\nFound: " + str(sorted(other[0].keys(), key=sortKey)))
			for h in self.headers():
				self.__headers[h] = DataColumn(self, h, list(self.__headers[h]) + [row[h] for row in other])
			self.__length = self.__length + len(other)
		elif isinstance(other, dict):
			if self.headers() and other and self.headers() != sorted(other.keys(), key=sortKey):
				raise DataTableException("headers don't match.  Expected: " + str(self.headers()) + "\nFound: " + str(sorted(other.keys(), key=sortKey)))
			if other:
				for h in self.headers():
					self.__headers[h] = DataColumn(self, h, list(self.__headers[h]) + [other[h]])
			self.__length = self.__length + 1
		else:
			print("other instance unknown: %s" % other.__class__)
			raise NotImplemented()
		return self
	append = __add__ = _copyAndApplyOp(__iadd__)
	def __isub__(self, other):
		'''remove the rows from other that are in self - uses exact match of rows'''
		if isinstance(other, dict):
			other = [other]
		indices = [i for i in range(len(self)) if self.getRow(i) not in other]
		for c in self.__headers.values():
			self.__headers[c.header] = DataColumn(self, c.header, [c[i] for i in indices])
		self.__length = len(indices)
		return self
	remove = __sub__ = _copyAndApplyOp(__isub__)
	def __iand__(self, other):
		'''Add columns to the data tabel using the dictionary keys from other as the new headers and their values as fields on each row
Overwrites existing columns'''
		if hasattr(other, '__call__'):
			if not self:
				return self
			results = []
			for row in self:
				results.append(other(row))
			for header in results[0].keys():
				self.__headers[header] = DataColumn(self, header, (row[header] for row in results))
			return self
		for header, value in other.items():
			data = []
			if hasattr(value, '__call__'):
				data = [value(row) for row in self]
			else:
				data = [value] * len(self)
			self.__headers[header] = DataColumn(self, header, data)
		return self
	extend = __and__ = _copyAndApplyOp(__iand__)
	def __or__(self, other):
		'''Pipes the DataTable into other
	Calls other with an iterator for the rows in self'''
		return other(iter(self))
	def __ixor__(self, other):
		'''remove column(s) from the data tabel'''
		if '__call__' in dir(other):
			for column in list(self.__headers.values()):
				if other(column):
					del self.__headers[column.header]
			return self
		if other in self.__headers:
			other = [other]
		for key in other:
			if key not in self.__headers:
				continue
			del self.__headers[key]
		return self
	exclude = __xor__ = _copyAndApplyOp(__ixor__)
	def __itruediv__(self, other):
		'''return new DataTable with only the columns listed in other'''
		if '__call__' in dir(other):
			for column in list(self.__headers.values()):
				if not other(column):
					del self.__headers[column.header]
			return self
		if other in self.__headers:
			other = [other]
		for key in list(self.__headers.keys()):
			if key in other:
				continue
			del self.__headers[key]
		return self
	project = __truediv__ = _copyAndApplyOp(__itruediv__)
	def removeBlankColumns(self):
		'''returns a copy of this DataTable with all of the blank columns removed'''
		blanks = [h for h, col in self.__headers.items() if not any(col)]
		return self.exclude(blanks)
	def sorted(self, *fields):
		'''returns a new copy of the data table sorted'''
		def key(idx):
			return tuple(sortKey(self.column(field)[idx]) for field in fields)
		idxs = sorted(range(len(self)), key=key)
		return self.select(idxs)
	def sort(self, *fields):
		other = self.sorted(*fields)
		for h in self.__headers.keys():
			self.__headers[h] = DataColumn(self, h, other.column(h))
		return self
	def iterBucket(self, *fields):
		buckets = defaultdict(lambda : [])
		for i in range(len(self)):
			key = tuple(self.__headers[field][i] for field in fields)
			buckets[key].append(i)
		for key, indices in buckets.items():
			yield key, self.select(indices)
	def sizeOfBuckets(self, *fields):
		'''Returns a dict of bucket -> number of items in the bucket'''
		buckets = defaultdict(lambda:0)
		for i in range(len(self)):
			key = tuple(self.__headers[field][i] for field in fields)
			buckets[key] += 1
		return buckets
	def bucket(self, *fields):
		'''Returns a dict of bucket -> DataTable of rows matching that bucket'''
		return dict(self.iterBucket(*fields))
	def filterBucket(self, predicate, *fields):
		'''Filter the datatable using an aggregate predicate
fields specifies how the data will be grouped
predicate is a method which takes a bucket of data and returns if the bucket should be included in the result
'''
		return DataTable.collect(bucket for key, bucket in self.iterBucket(*fields) if predicate(bucket))
	def join(self, other, joinParams=None,  otherFieldPrefix='',  joinType=JoinType.LEFT_OUTER_JOIN):
		'''
dataTable.join(otherTable, joinParams, otherFieldPrefix='')
	returns a new table with rows in the first table joined with rows in the second table, using joinParams to map fields in the first to fields in the second
Parameters:
	other - the table to join
	joinParams - a dictionary of <field in self> to <field in other>. Defaults to "natural join", merging common headers
	otherFieldPrefix - a string to prepend to the fields added from the second table
	joinType - the instance of JoinType which indicates if items should be included from one data table which aren't in the other
		'''
		if joinParams is None:
			joinParams = {h: h for h in self.headers() if h in other.headers()}
		if not isinstance(joinParams, dict):
			raise Exception("joinParams must be a dictionary of <field in self> to <field in other>")

		if not other:
			if not self or not joinType.leftOuter:
				return DataTable()
			return self & {otherFieldPrefix + v: None for v in other.headers() if v not in joinParams.values()}
		if not self:
			if not joinType.rightOuter:
				return DataTable()
			other = DataTable(other)
			for header in other.headers():
				if header not in joinParams.values():
					other.renameColumn(header, otherFieldPrefix + header)
			return other & {header: None for header in self.headers() if header not in joinParams}

		newHeaders = other.headers()
		for header in joinParams.values():
			newHeaders.remove(header)

		otherBuckets = other.bucket(*joinParams.values())
		def tempJoin():
			seenKeys = set()
			for row in self:
				newRow = dict(row)
				key = tuple(row[field] for field in joinParams.keys())
				seenKeys.add(key)
				if key not in otherBuckets:
					if joinType.leftOuter:
						for header in newHeaders:
							newRow[otherFieldPrefix+header] = None
						yield newRow
					continue
				otherRows = otherBuckets[key]
				for otherRow in otherRows:
					for header in newHeaders:
						newRow[otherFieldPrefix+header] = otherRow[header]
					yield dict(newRow)
			if joinType.rightOuter:
				joinParamReversed = {v: k for k,v in joinParams.items()}
				for otherRow in other:
					key = tuple(otherRow[field] for field in joinParams.values())
					if key not in seenKeys:
						newRow = {joinParamReversed[k] if k in joinParamReversed else otherFieldPrefix+k: v for k, v in otherRow.items()}
						for header in self.headers():
							if header not in joinParams:
								newRow[header] = None
						yield newRow
		return DataTable(tempJoin())
	def writeTo(self, fileName, *headers):
		'''Write the contents of this DataTable to a file with the given name in the standard csv format'''
		if not headers:
			headers = self.headers()
		with open(os.path.expanduser(fileName), 'w') as f:
			f.write(self | CSV_GivenHeaders(*headers))
	def duplicates(self, *fields):
		'''given a list of fields as keys, return a DataTable instance with the rows for which those fields are not unique'''
		matchCount = {}
		for i in range(len(self)):
			key = tuple(self.__headers[field][i] for field in fields)
			if key not in matchCount:
				matchCount[key] = 0
			else:
				matchCount[key] += 1
		return self.filter(lambda row: matchCount[tuple(row[field] for field in fields)])
	def _distinct(self):
		rows = set()
		for i, row in enumerate(zip(*self.__headers.values())):
			if row not in rows:
				yield i
				rows.add(row)
	def distinct(self):
		'''return a new DataTable with only unique rows'''
		return self.select(self._distinct())
	def fillDownBlanks(self, *fields):
		'''fills in the blanks in the current table such that each blank field in a row is filled in with the first non-blank entry in the column before it'''
		if not fields:
			fields = self.headers()
		for field in fields:
			self.__headers[field].fillDownBlanks()
	def pivot(self, rowID=None):
		'''Returns a new DataTable with the rows and columns swapped
In the resulting table, the headers from the previous table will be in the 'Field' column,
	then each row will be in the column Row0, Row1, ... RowN
	optional rowID is either a column to be used as the row identifier (fields in that column become new column headers),
or a method which takes a table (this table) and the row index and returns the column header corresponding with that row
		'''
		if rowID is None:
			digits = len(str(len(self)))
			fmt = 'Row%0' + str(digits) + 'd'
			getRowID = lambda dataTable, i: fmt % i
		elif isinstance(rowID, str):
			getRowID = lambda dataTable, i: dataTable[i][rowID]
		else:
			getRowID = rowID
		return DataTable([DataColumn(None, 'Field', self.headers())] +
						[DataColumn(None,
								getRowID(self, i),
								(self.__headers[h][i] for h in self.headers())) for i in range(len(self))])
	def aggregate(self, groupBy, aggregations={}):
		'''return an aggregation of the data grouped by a given set of fields.
Parameters:
	groupBy - the set of fields to group
	aggregations - a dict of field name -> aggregate method, where the method takes an intermediate DataTable
		and returns the value for that field for that row.
		'''
		if not aggregations:
			return self.project(groupBy).distinct()
		accumulatedRows = {}
		for row in self:
			key = tuple(row[field] for field in groupBy)
			if key not in accumulatedRows:
				accumulatedRows[key] = {a: agg.newBucket(row) for a, agg in aggregations.items()}
			accRow = accumulatedRows[key]
			for a, agg in aggregations.items():
				accRow[a] = agg.addRow(row, accRow[a])
		newData = []
		for key, accRow in sorted(accumulatedRows.items()):
			newData.append(AttributeDict(zip(groupBy, key)) + {a: agg.finalize(accRow[a]) for a, agg in aggregations.items()})
		return DataTable(newData)
	def renameColumn(self, column, newName):
		'''rename the column in place'''
		dataColumn = self.__headers[column]
		del self.__headers[column]
		self.__headers[newName] = dataColumn
		self.__headers[newName].header = newName
	def minRow(self, *fields):
		'''return the row with the minimum value(s) in the given field(s)'''
		if not self:
			return None
		return self.sorted(*fields)[0]
	def maxRow(self, *fields):
		'''return the row with the maximum value(s) in the given field(s)'''
		if not self:
			return None
		return self.sorted(*fields)[-1]

class DataRowProxy(object):
	def __init__(self, dataTable, idx):
		self.__dataTable = dataTable
		self.__idx = idx
	def __getitem__(self, header):
		if header not in self.keys():
			raise KeyError("Invalid key: %s" % header)
		return self.__dataTable.column(header)[self.__idx]
	def __setitem__(self, key, value):
		if key not in self.keys():
			raise KeyError("DataRowProxy doesn't support setting of keys not belonging to its associated DataTable")
		self.__dataTable.column(key)[self.__idx] = value
	def __contains__(self, key):
		return key in self.keys()
	def __getattr__(self, header):
		if header not in self.__dataTable.headers():
			raise AttributeError("Source table doesn't have column with header: " + header + ".  Available columns: %r" % self.__dataTable.headers())
		return self.__dataTable.column(header)[self.__idx]
	def __setattr__(self, key, value):
		if key.startswith('_') or key not in self.keys():
			super(DataRowProxy, self).__setattr__(key, value)
		else:
			self.__dataTable.column(key)[self.__idx] = value
	def __iter__(self):
		return ((column.header, column[self.__idx]) for column in self.__dataTable.columns())
	items = __iter__
	iteritems = __iter__
	def asDict(self):
		return AttributeDict(iter(self))
	def __str__(self):
		maxLen = max(len(str(h)) for h in self.keys())
		formatStr = '\n%' + str(maxLen) + 's : %s'
		return 'DataRowProxy: %s' % ','.join(formatStr % (header, value) for header, value in self)
	__repr__ = __str__
	def keys(self):
		return self.__dataTable.headers()
	def __add__(self, other):
		return self.asDict() + other