X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=PLC%2FFilter.py;h=d3a17b4e105cdd33208a087ddedac1787fbc3a23;hb=e70e20fdbececafef842ec7b330fd48db42e614e;hp=8a7ba310562832bdcf47c23b7f32bc817ab61610;hpb=f5cd72e35593c96b9996ab4e49674cccc7525c48;p=plcapi.git diff --git a/PLC/Filter.py b/PLC/Filter.py index 8a7ba31..d3a17b4 100644 --- a/PLC/Filter.py +++ b/PLC/Filter.py @@ -1,38 +1,33 @@ -# $Id$ -# $URL$ +# +# Thierry Parmentelat - INRIA +# from types import StringTypes -try: - set -except NameError: - from sets import Set - set = Set - import time from PLC.Faults import * from PLC.Parameter import Parameter, Mixed, python_type +from PLC.Logger import logger class Filter(Parameter, dict): """ A type of parameter that represents a filter on one or more columns of a database table. - Special features provide support for negation, upper and lower bounds, - as well as sorting and clipping. + Special features provide support for negation, upper and lower bounds, + sorting and clipping and more... - fields should be a dictionary of field names and types - Only filters on non-sequence type fields are supported. + fields should be a dictionary of field names and types. + As of PLCAPI-4.3-26, we provide support for filtering on + sequence types as well, with the special '&' and '|' modifiers. example : fields = {'node_id': Parameter(int, "Node identifier"), 'hostname': Parameter(int, "Fully qualified hostname", max = 255), ...} filter should be a dictionary of field names and values - representing the criteria for filtering. + representing the criteria for filtering. example : filter = { 'hostname' : '*.edu' , site_id : [34,54] } - Whether the filter represents an intersection (AND) or a union (OR) - of these criteria is determined by the join_with argument - provided to the sql method below + Special features: @@ -46,28 +41,61 @@ class Filter(Parameter, dict): example : filter = { '>time' : 1178531418 } in this example the integer value denotes a unix timestamp - * if a value is a sequence type, then it should represent + * if a value is a sequence type, then it should represent a list of possible values for that field example : filter = { 'node_id' : [12,34,56] } * a (string) value containing either a * or a % character is treated as a (sql) pattern; * are replaced with % that is the SQL wildcard character. - example : filter = { 'hostname' : '*.jp' } + example : filter = { 'hostname' : '*.jp' } - * fields starting with - are special and relate to row selection, i.e. sorting and clipping - * '-SORT' : a field name, or an ordered list of field names that are used for sorting - these fields may start with + (default) or - for denoting increasing or decreasing order + * a field starting with '&' or '|' should refer to a sequence type + the semantics is then that the object value (expected to be a list) + should contain all (&) or any (|) value specified in the corresponding + filter value. See other examples below. + example : filter = { '|role_ids' : [ 20, 40 ] } + example : filter = { '|roles' : ['tech', 'pi'] } + example : filter = { '&roles' : ['admin', 'tech'] } + example : filter = { '&roles' : 'tech' } + + * the filter's keys starting with '-' are special and relate to sorting and clipping + * '-SORT' : a field name, or an ordered list of field names that are used for sorting + these fields may start with + (default) or - for denoting increasing or decreasing order example : filter = { '-SORT' : [ '+node_id', '-hostname' ] } - * '-OFFSET' : the number of first rows to be ommitted - * '-LIMIT' : the amount of rows to be returned + * '-OFFSET' : the number of first rows to be ommitted + * '-LIMIT' : the amount of rows to be returned example : filter = { '-OFFSET' : 100, '-LIMIT':25} - A realistic example would read - GetNodes ( { 'node_type' : 'regular' , 'hostname' : '*.edu' , '-SORT' : 'hostname' , '-OFFSET' : 30 , '-LIMIT' : 25 } ) - and that would return regular (usual) nodes matching '*.edu' in alphabetical order from 31th to 55th + * similarly the two special keys below allow to change the semantics of multi-keys filters + * '-AND' : select rows that match ALL the criteria (default) + * '-OR' : select rows that match ANY criteria + The value attached to these keys is ignored. + Please note however that because a Filter is a dict, you cannot provide two criteria on a given key. + + + Here are a few realistic examples + + GetNodes ( { 'node_type' : 'regular' , 'hostname' : '*.edu' , + '-SORT' : 'hostname' , '-OFFSET' : 30 , '-LIMIT' : 25 } ) + would return regular (usual) nodes matching '*.edu' in alphabetical order from 31th to 55th + + GetNodes ( { '~peer_id' : None } ) + returns the foreign nodes - that have an integer peer_id + + GetPersons ( { '|role_ids' : [ 20 , 40] } ) + would return all persons that have either pi (20) or tech (40) roles + + GetPersons ( { '&role_ids' : 10 } ) + GetPersons ( { '&role_ids' : 10 } ) + GetPersons ( { '|role_ids' : [ 10 ] } ) + GetPersons ( { '|role_ids' : [ 10 ] } ) + all 4 forms are equivalent and would return all admin users in the system """ + debug=False +# debug=True + def __init__(self, fields = {}, filter = {}, doc = "Attribute filter"): # Store the filter in our dict instance dict.__init__(self, filter) @@ -75,9 +103,8 @@ class Filter(Parameter, dict): # Declare ourselves as a type of parameter that can take # either a value or a list of values for each of the specified # fields. - self.fields = dict ( [ ( field, Mixed (expected, [expected])) - for (field,expected) in fields.iteritems() - if python_type(expected) not in (list, tuple, set) ] ) + self.fields = dict ( [ ( field, Mixed (expected, [expected])) + for (field,expected) in fields.iteritems() ] ) # Null filter means no filter Parameter.__init__(self, self.fields, doc = doc, nullok = True) @@ -87,6 +114,15 @@ class Filter(Parameter, dict): Returns a SQL conditional that represents this filter. """ + if self.has_key('-AND'): + del self['-AND'] + join_with='AND' + if self.has_key('-OR'): + del self['-OR'] + join_with='OR' + + self.join_with=join_with + # So that we always return something if join_with == "AND": conditionals = ["True"] @@ -95,51 +131,45 @@ class Filter(Parameter, dict): else: assert join_with in ("AND", "OR") - # init + # init sorts = [] clips = [] for field, value in self.iteritems(): - # handle negation, numeric comparisons - # simple, 1-depth only mechanism + # handle negation, numeric comparisons + # simple, 1-depth only mechanism - modifiers={'~' : False, - '<' : False, '>' : False, - '[' : False, ']' : False, + modifiers={'~' : False, + '<' : False, '>' : False, + '[' : False, ']' : False, '-' : False, - } - - for char in modifiers.keys(): - if field[0] == char: - modifiers[char]=True; - field = field[1:] - break + '&' : False, '|' : False, + } + def check_modifiers(field): + if field[0] in modifiers.keys(): + modifiers[field[0]] = True + field = field[1:] + return check_modifiers(field) + return field + field = check_modifiers(field) # filter on fields if not modifiers['-']: if field not in self.fields: raise PLCInvalidArgument, "Invalid filter field '%s'" % field - if isinstance(value, (list, tuple, set)): - # handling filters like '~slice_id':[] - # this should return true, as it's the opposite of 'slice_id':[] which is false - # prior to this fix, 'slice_id':[] would have returned ``slice_id IN (NULL) '' which is unknown - # so it worked by coincidence, but the negation '~slice_ids':[] would return false too - if not value: - field="" - operator="" - value = "FALSE" - else: - operator = "IN" - value = map(str, map(api.db.quote, value)) - value = "(%s)" % ", ".join(value) - else: + # handling array fileds always as compound values + if modifiers['&'] or modifiers['|']: + if not isinstance(value, (list, tuple, set)): + value = [value,] + + def get_op_and_val(value): if value is None: operator = "IS" value = "NULL" elif isinstance(value, StringTypes) and \ (value.find("*") > -1 or value.find("%") > -1): - operator = "LIKE" + operator = "ILIKE" # insert *** in pattern instead of either * or % # we dont use % as requests are likely to %-expansion later on # actual replacement to % done in PostgreSQL.py @@ -156,10 +186,50 @@ class Filter(Parameter, dict): operator='<=' if modifiers[']']: operator='>=' + value = str(api.db.quote(value)) + return (operator, value) + + if isinstance(value, (list, tuple, set)): + # handling filters like '~slice_id':[] + # this should return true, as it's the opposite of 'slice_id':[] which is false + # prior to this fix, 'slice_id':[] would have returned ``slice_id IN (NULL) '' which is unknown + # so it worked by coincidence, but the negation '~slice_ids':[] would return false too + if not value: + if modifiers['&'] or modifiers['|']: + operator = "=" + value = "'{}'" else: - value = str(api.db.quote(value)) - - clause = "%s %s %s" % (field, operator, value) + field="" + operator="" + value = "FALSE" + clause = "%s %s %s" % (field, operator, value) + else: + vals = {} + for val in value: + base_op, val = get_op_and_val(val) + if base_op in vals: + vals[base_op].append(val) + else: + vals[base_op] = [val] + subclauses = [] + for operator in vals.keys(): + if operator == '=': + if modifiers['&']: + subclauses.append("(%s @> ARRAY[%s])" % (field, ",".join(vals[operator]))) + elif modifiers['|']: + subclauses.append("(%s && ARRAY[%s])" % (field, ",".join(vals[operator]))) + else: + subclauses.append("(%s IN (%s))" % (field, ",".join(vals[operator]))) + elif operator == 'IS': + subclauses.append("(%s IS NULL)" % field) + else: + for value in vals[operator]: + subclauses.append("(%s %s %s)" % (field, operator, value)) + clause = "(" + " OR ".join(subclauses) + ")" + else: + operator, value = get_op_and_val(value) + + clause = "%s %s %s" % (field, operator, value) if modifiers['~']: clause = " ( NOT %s ) " % (clause) @@ -196,5 +266,7 @@ class Filter(Parameter, dict): clip_part += " ORDER BY " + ",".join(sorts) if clips: clip_part += " " + " ".join(clips) -# print 'where_part=',where_part,'clip_part',clip_part - return (where_part,clip_part) + if Filter.debug: + logger.debug('Filter.sql: where_part={} - clip_part={}' + .format(where_part, clip_part)) + return where_part, clip_part