def format_cube(decoders, aggs, start, query, select):
# decoders = sorted(decoders, key=lambda d: -d.edge.dim) # REVERSE DECODER ORDER, BECAUSE ES QUERY WAS BUILT IN REVERSE ORDER
new_edges = count_dim(aggs, decoders)
dims = []
for e in new_edges:
if isinstance(e.value, TupleOp):
e.allowNulls = False
extra = 0 if e.allowNulls is False else 1
dims.append(len(e.domain.partitions) + extra)
dims = tuple(dims)
matricies = [(s, Matrix(dims=dims, zeros=s.default)) for s in select]
for row, coord, agg in aggs_iterator(aggs, decoders):
for s, m in matricies:
try:
v = s.pull(agg)
m[coord] = v
except Exception as e:
# THIS HAPPENS WHEN ES RETURNS MORE TUPLE COMBINATIONS THAN DOCUMENTS
if agg.get('doc_count') != 0:
Log.error("Programmer error", cause=e)
cube = Cube(
query.select,
sort_using_key(new_edges, key=lambda e: e.dim), # ENSURE EDGES ARE IN SAME ORDER AS QUERY
{s.name: m for s, m in matricies}
)
cube.frum = query
return cube
def format_cube_from_aggop(decoders, aggs, start, query, select):
agg = drill(aggs)
matricies = [(s, Matrix(dims=[], zeros=s.default)) for s in select]
for s, m in matricies:
m[tuple()] = s.pull(agg)
cube = Cube(query.select, [], {s.name: m for s, m in matricies})
cube.frum = query
return cube
def _es_terms2(es, mvel, query):
"""
WE ASSUME THERE ARE JUST TWO EDGES, AND EACH HAS A SIMPLE value
"""
# REQUEST VALUES IN FIRST DIMENSION
q1 = query.copy()
q1.edges = query.edges[0:1:]
values1 = es_terms(es, mvel, q1).edges[0].domain.partitions.value
select = listwrap(query.select)
FromES = build_es_query(query)
for s in select:
for i, v in enumerate(values1):
FromES.facets[s.name + "," + str(i)] = {
"terms": {
"field": query.edges[1].value,
"size": coalesce(query.limit, 200000)
},
"facet_filter": simplify_esfilter({"and": [
query.where,
{"term": {query.edges[0].value: v}}
]})
}
data = es_post(es, FromES, query.limit)
# UNION ALL TERMS FROM SECOND DIMENSION
values2 = set()
for k, f in data.facets.items():
values2.update(f.terms.term)
values2 = jx.sort(values2)
term2index = {v: i for i, v in enumerate(values2)}
query.edges[1].domain.partitions = FlatList([{"name": v, "value": v} for v in values2])
# MAKE CUBE
output = {}
dims = [len(values1), len(values2)]
for s in select:
output[s.name] = Matrix(*dims)
# FILL CUBE
# EXPECTING ONLY SELECT CLAUSE FACETS
for facetName, facet in data.facets.items():
coord = facetName.split(",")
s = [s for s in select if s.name == coord[0]][0]
i1 = int(coord[1])
for term in facet.terms:
i2 = term2index[term.term]
output[s.name][(i1, i2)] = term[aggregates[s.aggregate]]
cube = Cube(query.select, query.edges, output)
cube.query = query
return cube
def format_cube(T, select, query=None):
with Timer("format table"):
table = format_table(T, select, query)
if len(table.data) == 0:
return Cube(
scrub_select(select),
edges=[{"name": "rownum", "domain": {"type": "rownum", "min": 0, "max": 0, "interval": 1}}],
data={h: Matrix(list=[]) for i, h in enumerate(table.header)}
)
cols = transpose(*unwrap(table.data))
return Cube(
scrub_select(select),
edges=[{"name": "rownum", "domain": {"type": "rownum", "min": 0, "max": len(table.data), "interval": 1}}],
data={h: Matrix(list=cols[i]) for i, h in enumerate(table.header)}
)
def es_deepop(es, mvel, query):
FromES = es09.util.build_es_query(query)
select = query.edges
temp_query = query.copy()
temp_query.select = select
temp_query.edges = FlatList()
FromES.facets.mvel = {
"terms": {
"script_field": mvel.code(temp_query),
"size": query.limit
},
"facet_filter": jx_expression(query.where).to_esfilter()
}
data = es_post(es, FromES, query.limit)
rows = unpack_terms(data.facets.mvel, query.edges)
terms = transpose(*rows)
# NUMBER ALL EDGES FOR JSON EXPRESSION INDEXING
edges = query.edges
for f, e in enumerate(edges):
for r in terms[f]:
e.domain.getPartByKey(r)
e.index = f
for p, part in enumerate(e.domain.partitions):
part.dataIndex = p
e.domain.NULL.dataIndex = len(e.domain.partitions)
# MAKE CUBE
dims = [len(e.domain.partitions) for e in query.edges]
output = Matrix(*dims)
# FILL CUBE
for r in rows:
term_coord = [
e.domain.getPartByKey(r[i]).dataIndex for i, e in enumerate(edges)
]
output[term_coord] = SUM(output[term_coord], r[-1])
cube = Cube(query.select, query.edges, {query.select.name: output})
cube.frum = query
return cube
def format_cube(aggs, es_query, query, decoders, all_selects):
new_edges = count_dim(aggs, es_query, decoders)
dims = []
for e in new_edges:
if is_op(e.value, TupleOp):
e.allowNulls = False
extra = 0 if e.allowNulls is False else 1
dims.append(len(e.domain.partitions) + extra)
dims = tuple(dims)
if any(s.default != canonical_aggregates[s.aggregate].default for s in all_selects):
# UNUSUAL DEFAULT VALUES MESS THE union() FUNCTION
is_default = Matrix(dims=dims, zeros=True)
matricies = {s.name: Matrix(dims=dims) for s in all_selects}
for row, coord, agg, selects in aggs_iterator(aggs, es_query, decoders):
for select in selects:
m = matricies[select.name]
v = select.pull(agg)
if v == None:
continue
is_default[coord] = False
union(m, coord, v, select.aggregate)
# FILL THE DEFAULT VALUES
for c, v in is_default:
if v:
for s in all_selects:
matricies[s.name][c] = s.default
else:
matricies = {s.name: Matrix(dims=dims, zeros=s.default) for s in all_selects}
for row, coord, agg, selects in aggs_iterator(aggs, es_query, decoders):
for select in selects:
m = matricies[select.name]
v = select.pull(agg)
union(m, coord, v, select.aggregate)
cube = Cube(
query.select,
sort_using_key(new_edges, key=lambda e: e.dim), # ENSURE EDGES ARE IN SAME ORDER AS QUERY
matricies
)
cube.frum = query
return cube
def format_cube(aggs, es_query, query, decoders, all_selects):
new_edges = count_dim(aggs, es_query, decoders)
dims = []
for e in new_edges:
if is_op(e.value, TupleOp):
e.allowNulls = False
extra = 0 if e.allowNulls is False else 1
dims.append(len(e.domain.partitions) + extra)
dims = tuple(dims)
if any(s.default != canonical_aggregates[s.aggregate].default for s in all_selects):
# UNUSUAL DEFAULT VALUES MESS THE union() FUNCTION
is_default = Matrix(dims=dims, zeros=True)
matricies = {s.name: Matrix(dims=dims) for s in all_selects}
for row, coord, agg, selects in aggs_iterator(aggs, es_query, decoders):
for select in selects:
m = matricies[select.name]
v = select.pull(agg)
if v == None:
continue
is_default[coord] = False
union(m, coord, v, select.aggregate)
# FILL THE DEFAULT VALUES
for c, v in is_default:
if v:
for s in all_selects:
matricies[s.name][c] = s.default
else:
matricies = {s.name: Matrix(dims=dims, zeros=s.default) for s in all_selects}
for row, coord, agg, selects in aggs_iterator(aggs, es_query, decoders):
for select in selects:
m = matricies[select.name]
v = select.pull(agg)
union(m, coord, v, select.aggregate)
cube = Cube(
query.select,
sort_using_key(new_edges, key=lambda e: e.dim), # ENSURE EDGES ARE IN SAME ORDER AS QUERY
matricies
)
cube.frum = query
return cube
def es_countop(es, mvel, query):
"""
RETURN SINGLE COUNT
"""
select = listwrap(query.select)
FromES = build_es_query(query)
for s in select:
if is_variable_name(s.value):
FromES.facets[s.name] = {
"terms": {
"field": s.value,
"size": query.limit,
},
"facet_filter": {
"exists": {
"field": s.value
}
}
}
else:
# COMPLICATED value IS PROBABLY A SCRIPT, USE IT
FromES.facets[s.name] = {
"terms": {
"script_field":
es09.expressions.compile_expression(s.value, query),
"size":
200000
}
}
data = es_post(es, FromES, query.limit)
matricies = {}
for s in select:
matricies[s.name] = Matrix(value=data.hits.facets[s.name].total)
cube = Cube(query.select, query.edges, matricies)
cube.frum = query
return cube
def es_aggop(es, mvel, query):
select = listwrap(query.select)
FromES = build_es_query(query)
isSimple = AND(aggregates[s.aggregate] == "count" for s in select)
if isSimple:
return es_countop(es, query) # SIMPLE, USE TERMS FACET INSTEAD
value2facet = dict() # ONLY ONE FACET NEEDED PER
name2facet = dict() # MAP name TO FACET WITH STATS
for s in select:
if s.value not in value2facet:
if isinstance(s.value, Variable):
unwrap(FromES.facets)[s.name] = {
"statistical": {
"field": s.value.var
},
"facet_filter": query.where.to_esfilter()
}
else:
unwrap(FromES.facets)[s.name] = {
"statistical": {
"script": jx_expression_to_function(s.value)
},
"facet_filter": query.where.to_es_filter()
}
value2facet[s.value] = s.name
name2facet[s.name] = value2facet[s.value]
data = es_post(es, FromES, query.limit)
matricies = {
s.name: Matrix(value=fix_es_stats(data.facets[literal_field(s.name)])[
aggregates[s.aggregate]])
for s in select
}
cube = Cube(query.select, [], matricies)
cube.frum = query
return cube
def save(self, query):
query.meta = None
json = convert.value2json(query)
hash = unicode2utf8(json)
# TRY MANY HASHES AT ONCE
hashes = [None] * HASH_BLOCK_SIZE
for i in range(HASH_BLOCK_SIZE):
hash = hashlib.sha1(hash).digest()
hashes[i] = hash
short_hashes = [
convert.bytes2base64(h[0:6]).replace("/", "_") for h in hashes
]
available = {h: True for h in short_hashes}
existing = self.es.query({
"from": "saved_queries",
"where": {
"terms": {
"hash": short_hashes
}
},
"meta": {
"timeout": "2second"
}
})
for e in Cube(select=existing.select,
edges=existing.edges,
data=existing.data).values():
if e.query == json:
return e.hash
available[e.hash] = False
# THIS WILL THROW AN ERROR IF THERE ARE NONE, HOW UNLUCKY!
best = [h for h in short_hashes if available[h]][0]
self.queue.add({
"id": best,
"value": {
"hash": best,
"create_time": Date.now(),
"last_used": Date.now(),
"query": json
}
})
Log.note("Saved query as {{hash}}", hash=best)
return best
def list_aggs(frum, query):
frum = wrap(frum)
select = listwrap(query.select)
for e in query.edges:
if isinstance(e.domain, DefaultDomain):
accessor = jx_expression_to_function(e.value)
unique_values = set(map(accessor, frum))
if None in unique_values:
e.allowNulls = coalesce(e.allowNulls, True)
unique_values -= {None}
e.domain = SimpleSetDomain(partitions=list(sorted(unique_values)))
else:
pass
s_accessors = [(ss.name, compile_expression(ss.value.to_python()))
for ss in select]
result = {
s.name: Matrix(dims=[
len(e.domain.partitions) + (1 if e.allowNulls else 0)
for e in query.edges
],
zeros=lambda: windows.name2accumulator.get(s.aggregate)
(**s))
for s in select
}
where = jx_expression_to_function(query.where)
coord = [None] * len(query.edges)
edge_accessor = [(i, make_accessor(e)) for i, e in enumerate(query.edges)]
net_new_edge_names = set(wrap(query.edges).name) - UNION(
e.value.vars() for e in query.edges)
if net_new_edge_names & UNION(ss.value.vars() for ss in select):
# s_accessor NEEDS THESE EDGES, SO WE PASS THEM ANYWAY
for d in filter(where, frum):
d = d.copy()
for c, get_matches in edge_accessor:
coord[c] = get_matches(d)
for s_name, s_accessor in s_accessors:
mat = result[s_name]
for c in itertools.product(*coord):
acc = mat[c]
for e, cc in zip(query.edges, c):
d[e.name] = e.domain.partitions[cc]
val = s_accessor(d, c, frum)
acc.add(val)
else:
# FASTER
for d in filter(where, frum):
for c, get_matches in edge_accessor:
coord[c] = get_matches(d)
for s_name, s_accessor in s_accessors:
mat = result[s_name]
for c in itertools.product(*coord):
acc = mat[c]
val = s_accessor(d, c, frum)
acc.add(val)
for s in select:
# if s.aggregate == "count":
# continue
m = result[s.name]
for c, var in m.items():
if var != None:
m[c] = var.end()
from jx_python.containers.cube import Cube
output = Cube(select, query.edges, result)
return output
def save(self, query):
"""
SAVE query TO ES FOR LATER RECOVERY
:param query:
:return: HAS TO USE FOR RECOVERY
"""
meta, query.meta = query.meta, None
json = convert.value2json(query)
query.meta = meta
hash = json.encode("utf8")
# TRY MANY HASHES AT ONCE
hashes = [None] * HASH_BLOCK_SIZE
for i in range(HASH_BLOCK_SIZE):
hash = hashlib.sha1(hash).digest()
hashes[i] = hash
short_hashes = [convert.bytes2base64(h[0:6]).replace("/", "_") for h in hashes]
available = {h: True for h in short_hashes}
existing = self.es.query(
{
"from": "saved_queries",
"where": {"terms": {"hash": short_hashes}},
"meta": {"timeout": "2second"},
}
)
for e in Cube(
select=existing.select, edges=existing.edges, data=existing.data
).values():
if e.query == json:
return e.hash
available[e.hash] = False
# THIS WILL THROW AN ERROR IF THERE ARE NONE, HOW UNLUCKY!
best = first(h for h in short_hashes if available[h])
self.queue.add(
{
"id": best,
"value": {
"hash": best,
"create_time": Date.now(),
"last_used": Date.now(),
"query": json,
},
}
)
if meta.testing:
while True:
verify = self.es.query(
{
"from": "saved_queries",
"where": {"terms": {"hash": best}},
"meta": {"timeout": "2second"},
"select": "query",
"format": "list",
}
)
if verify.data:
break
Log.alert("wait for saved query")
Till(seconds=1).wait()
Log.note("Saved {{json}} query as {{hash}}", json=json, hash=best)
return best
def es_terms(es, mvel, query):
"""
RETURN LIST OF ALL EDGE QUERIES
EVERY FACET IS NAMED <select.name>, <c1>, ... <cN> WHERE <ci> ARE THE ELEMENT COORDINATES
WE TRY TO PACK DIMENSIONS INTO THE TERMS TO MINIMIZE THE CROSS-PRODUCT EXPLOSION
"""
if len(query.edges) == 2:
return _es_terms2(es, mvel, query)
select = listwrap(query.select)
FromES = build_es_query(query)
packed_term = compileEdges2Term(mvel, query.edges, wrap([]))
for s in select:
FromES.facets[s.name] = {
"terms": {
"field": packed_term.field,
"script_field": packed_term.expression,
"size": coalesce(query.limit, 200000)
},
"facet_filter": simplify_esfilter(query.where)
}
term2Parts = packed_term.term2parts
data = es_post(es, FromES, query.limit)
# GETTING ALL PARTS WILL EXPAND THE EDGES' DOMAINS
# BUT HOW TO UNPACK IT FROM THE term FASTER IS UNKNOWN
for k, f in data.facets.items():
for t in f.terms:
term2Parts(t.term)
# NUMBER ALL EDGES FOR jx INDEXING
for f, e in enumerate(query.edges):
e.index = f
if e.domain.type in ["uid", "default"]:
# e.domain.partitions = jx.sort(e.domain.partitions, "value")
for p, part in enumerate(e.domain.partitions):
part.dataIndex = p
e.domain.NULL.dataIndex = len(e.domain.partitions)
# MAKE CUBE
output = {}
dims = [len(e.domain.partitions) + (1 if e.allowNulls else 0) for e in query.edges]
for s in select:
output[s.name] = Matrix(*dims)
# FILL CUBE
# EXPECTING ONLY SELECT CLAUSE FACETS
for facetName, facet in data.facets.items():
for term in facet.terms:
term_coord = term2Parts(term.term).dataIndex
for s in select:
try:
output[s.name][term_coord] = term[aggregates[s.aggregate]]
except Exception as e:
# USUALLY CAUSED BY output[s.name] NOT BEING BIG ENOUGH TO HANDLE NULL COUNTS
pass
cube = Cube(query.select, query.edges, output)
cube.query = query
return cube
def es_terms_stats(esq, mvel, query):
select = listwrap(query.select)
facetEdges = [] # EDGES THAT WILL REQUIRE A FACET FOR EACH PART
termsEdges = FlatList()
specialEdge = None
special_index = -1
# A SPECIAL EDGE IS ONE THAT HAS AN UNDEFINED NUMBER OF PARTITIONS AT QUERY TIME
# FIND THE specialEdge, IF ONE
for f, tedge in enumerate(query.edges):
if tedge.domain.type in domains.KNOWN:
for p, part in enumerate(tedge.domain.partitions):
part.dataIndex = p
# FACETS ARE ONLY REQUIRED IF SQL JOIN ON DOMAIN IS REQUIRED (RANGE QUERY)
# OR IF WE ARE NOT SIMPLY COUNTING
# OR IF NO SCRIPTING IS ALLOWED (SOME OTHER CODE IS RESPONSIBLE FOR SETTING isFacet)
# OR IF WE JUST WANT TO FORCE IT :)
# OF COURSE THE default EDGE IS NOT EXPLICIT, SO MUST BE A TERM
facetEdges.append(tedge)
else:
if specialEdge:
Log.error(
"There is more than one open-ended edge: self can not be handled"
)
specialEdge = tedge
special_index = f
termsEdges.append(tedge)
if not specialEdge:
# WE SERIOUSLY WANT A SPECIAL EDGE, OTHERWISE WE WILL HAVE TOO MANY FACETS
# THE BIGGEST EDGE MAY BE COLLAPSED TO A TERM, MAYBE?
num_parts = 0
special_index = -1
for i, e in enumerate(facetEdges):
l = len(e.domain.partitions)
if ((e.value and is_variable_name(e.value))
or len(e.domain.dimension.fields) == 1) and l > num_parts:
num_parts = l
specialEdge = e
special_index = i
facetEdges.pop(special_index)
termsEdges.append(specialEdge)
total_facets = PRODUCT(len(f.domain.partitions)
for f in facetEdges) * len(select)
if total_facets > 100:
# WE GOT A PROBLEM, LETS COUNT THE SIZE OF REALITY:
counts = esq.query({
"from": query.frum,
"select": {
"aggregate": "count"
},
"edges": facetEdges,
"where": query.where,
"limit": query.limit
})
esFacets = []
def add_facet(value, parts, cube):
if value:
esFacets.append(parts)
counts["count"].forall(add_facet)
Log.note(
"{{theory_count}} theoretical combinations, {{real_count}} actual combos found",
real_count=len(esFacets),
theory_count=total_facets)
if not esFacets:
# MAKE EMPTY CUBE
matricies = {}
dims = [
len(e.domain.partitions) + (1 if e.allowNulls else 0)
for e in query.edges
]
for s in select:
matricies[s.name] = Matrix(*dims)
cube = Cube(query.select, query.edges, matricies)
cube.frum = query
return cube
else:
# GENERATE ALL COMBOS
esFacets = getAllEdges(facetEdges)
calcTerm = compileEdges2Term(mvel, termsEdges, FlatList())
term2parts = calcTerm.term2parts
if len(esFacets) * len(select) > 1000:
Log.error(
"not implemented yet"
) # WE HAVE SOME SERIOUS PERMUTATIONS, WE MUST ISSUE MULTIPLE QUERIES
pass
FromES = build_es_query(query)
for s in select:
for parts in esFacets:
condition = FlatList()
constants = FlatList()
name = [literal_field(s.name)]
for f, fedge in enumerate(facetEdges):
name.append(str(parts[f].dataIndex))
condition.append(buildCondition(mvel, fedge, parts[f]))
constants.append({
"name": fedge.domain.name,
"value": parts[f]
})
condition.append(query.where)
name = ",".join(name)
FromES.facets[name] = {
"terms_stats": {
"key_field":
calcTerm.field,
"value_field":
s.value if is_variable_name(s.value) else None,
"value_script":
mvel.compile_expression(s.value)
if not is_variable_name(s.value) else None,
"size":
coalesce(query.limit, 200000)
}
}
if condition:
FromES.facets[name].facet_filter = simplify_esfilter(
{"and": condition})
data = es_post(esq.es, FromES, query.limit)
if specialEdge.domain.type not in domains.KNOWN:
# WE BUILD THE PARTS BASED ON THE RESULTS WE RECEIVED
partitions = FlatList()
map = {}
for facetName, parts in data.facets.items():
for stats in parts.terms:
if not map[stats]:
part = {"value": stats, "name": stats}
partitions.append(part)
map[stats] = part
partitions.sort(specialEdge.domain.compare)
for p, part in enumerate(partitions):
part.dataIndex = p
specialEdge.domain.map = map
specialEdge.domain.partitions = partitions
# MAKE CUBE
matricies = {}
dims = [
len(e.domain.partitions) + (1 if e.allowNulls else 0)
for e in query.edges
]
for s in select:
matricies[s.name] = Matrix(*dims)
name2agg = {s.name: aggregates[s.aggregate] for s in select}
# FILL CUBE
for edgeName, parts in data.facets.items():
temp = edgeName.split(",")
pre_coord = tuple(int(c) for c in temp[1:])
sname = temp[0]
for stats in parts.terms:
if specialEdge:
special = term2parts(stats.term)[0]
coord = pre_coord[:special_index] + (
special.dataIndex, ) + pre_coord[special_index:]
else:
coord = pre_coord
matricies[sname][coord] = stats[name2agg[sname]]
cube = Cube(query.select, query.edges, matricies)
cube.frum = query
return cube
def _es_terms2(es, mvel, query):
"""
WE ASSUME THERE ARE JUST TWO EDGES, AND EACH HAS A SIMPLE value
"""
# REQUEST VALUES IN FIRST DIMENSION
q1 = query.copy()
q1.edges = query.edges[0:1:]
values1 = es_terms(es, mvel, q1).edges[0].domain.partitions.value
select = listwrap(query.select)
FromES = build_es_query(query)
for s in select:
for i, v in enumerate(values1):
FromES.facets[s.name + "," + str(i)] = {
"terms": {
"field": query.edges[1].value,
"size": coalesce(query.limit, 200000)
},
"facet_filter":
simplify_esfilter({
"and": [query.where, {
"term": {
query.edges[0].value: v
}
}]
})
}
data = es_post(es, FromES, query.limit)
# UNION ALL TERMS FROM SECOND DIMENSION
values2 = set()
for k, f in data.facets.items():
values2.update(f.terms.term)
values2 = jx.sort(values2)
term2index = {v: i for i, v in enumerate(values2)}
query.edges[1].domain.partitions = FlatList([{
"name": v,
"value": v
} for v in values2])
# MAKE CUBE
output = {}
dims = [len(values1), len(values2)]
for s in select:
output[s.name] = Matrix(*dims)
# FILL CUBE
# EXPECTING ONLY SELECT CLAUSE FACETS
for facetName, facet in data.facets.items():
coord = facetName.split(",")
s = [s for s in select if s.name == coord[0]][0]
i1 = int(coord[1])
for term in facet.terms:
i2 = term2index[term.term]
output[s.name][(i1, i2)] = term[aggregates[s.aggregate]]
cube = Cube(query.select, query.edges, output)
cube.query = query
return cube
def es_setop(es, mvel, query):
FromES = es09.util.build_es_query(query)
select = listwrap(query.select)
isDeep = len(split_field(
query.frum.name)) > 1 # LOOKING INTO NESTED WILL REQUIRE A SCRIPT
isComplex = OR([
s.value == None and s.aggregate not in ("count", "none")
for s in select
]) # CONVERTING esfilter DEFINED PARTS WILL REQUIRE SCRIPT
if not isDeep and not isComplex:
if len(select) == 1 and isinstance(select[0].value, LeavesOp):
FromES = wrap({
"query": {
"bool": {
"query": {
"match_all": {}
},
"filter": query.where.to_esfilter()
}
},
"sort": query.sort,
"size": 0
})
elif all(isinstance(v, Variable) for v in select.value):
FromES = wrap({
"query": {
"bool": {
"query": {
"match_all": {}
},
"filter": query.where.to_esfilter()
}
},
"fields": select.value,
"sort": query.sort,
"size": coalesce(query.limit, 200000)
})
elif not isDeep:
simple_query = query.copy()
simple_query.where = TRUE # THE FACET FILTER IS FASTER
FromES.facets.mvel = {
"terms": {
"script_field": mvel.code(simple_query),
"size": coalesce(simple_query.limit, 200000)
},
"facet_filter": jx_expression(query.where).to_esfilter()
}
else:
FromES.facets.mvel = {
"terms": {
"script_field": mvel.code(query),
"size": coalesce(query.limit, 200000)
},
"facet_filter": jx_expression(query.where).to_esfilter()
}
data = es_post(es, FromES, query.limit)
if len(select) == 1 and isinstance(select[0].value, LeavesOp):
# SPECIAL CASE FOR SINGLE COUNT
cube = wrap(data).hits.hits._source
elif isinstance(select[0].value, Variable):
# SPECIAL CASE FOR SINGLE TERM
cube = wrap(data).hits.hits.fields
else:
data_list = unpack_terms(data.facets.mvel, select)
if not data_list:
cube = Cube(select, [], {s.name: Matrix.wrap([]) for s in select})
else:
output = transpose(*data_list)
cube = Cube(
select, [],
{s.name: Matrix(list=output[i])
for i, s in enumerate(select)})
return Data(meta={"esquery": FromES}, data=cube)
def es_terms(es, mvel, query):
"""
RETURN LIST OF ALL EDGE QUERIES
EVERY FACET IS NAMED <select.name>, <c1>, ... <cN> WHERE <ci> ARE THE ELEMENT COORDINATES
WE TRY TO PACK DIMENSIONS INTO THE TERMS TO MINIMIZE THE CROSS-PRODUCT EXPLOSION
"""
if len(query.edges) == 2:
return _es_terms2(es, mvel, query)
select = listwrap(query.select)
FromES = build_es_query(query)
packed_term = compileEdges2Term(mvel, query.edges, wrap([]))
for s in select:
FromES.facets[s.name] = {
"terms": {
"field": packed_term.field,
"script_field": packed_term.expression,
"size": coalesce(query.limit, 200000)
},
"facet_filter": simplify_esfilter(query.where)
}
term2Parts = packed_term.term2parts
data = es_post(es, FromES, query.limit)
# GETTING ALL PARTS WILL EXPAND THE EDGES' DOMAINS
# BUT HOW TO UNPACK IT FROM THE term FASTER IS UNKNOWN
for k, f in data.facets.items():
for t in f.terms:
term2Parts(t.term)
# NUMBER ALL EDGES FOR jx INDEXING
for f, e in enumerate(query.edges):
e.index = f
if e.domain.type in ["uid", "default"]:
# e.domain.partitions = jx.sort(e.domain.partitions, "value")
for p, part in enumerate(e.domain.partitions):
part.dataIndex = p
e.domain.NULL.dataIndex = len(e.domain.partitions)
# MAKE CUBE
output = {}
dims = [
len(e.domain.partitions) + (1 if e.allowNulls else 0)
for e in query.edges
]
for s in select:
output[s.name] = Matrix(*dims)
# FILL CUBE
# EXPECTING ONLY SELECT CLAUSE FACETS
for facetName, facet in data.facets.items():
for term in facet.terms:
term_coord = term2Parts(term.term).dataIndex
for s in select:
try:
output[s.name][term_coord] = term[aggregates[s.aggregate]]
except Exception as e:
# USUALLY CAUSED BY output[s.name] NOT BEING BIG ENOUGH TO HANDLE NULL COUNTS
pass
cube = Cube(query.select, query.edges, output)
cube.query = query
return cube
def cube_aggs(frum, query):
select = listwrap(query.select)
#MATCH EDGES IN QUERY TO ONES IN frum
for e in query.edges:
for fs in frum.select:
if fs.name == e.value:
Log.error("Not implemented yet")
if isinstance(e.domain, DefaultDomain):
# DEFAULT DOMAINS CAN EASILY BE LOOKED UP FROM frum
for fe in frum.edges:
if fe.name == e.value:
e.domain = SimpleSetDomain(**fe.domain.__data__())
e.value = e.value + "." + fe.domain.key
break
else:
for fe in frum.edges:
if fe.name == e.value:
e.value = e.value + "." + fe.domain.key
break
result = {
s.name: Matrix(dims=[
len(e.domain.partitions) + (1 if e.allowNulls else 0)
for e in query.edges
],
zeros=s.default)
for s in select
}
where = jx_expression_to_function(query.where)
for d in filter(where, frum.values()):
coord = [
] # LIST OF MATCHING COORDINATE FAMILIES, USUALLY ONLY ONE PER FAMILY BUT JOINS WITH EDGES CAN CAUSE MORE
for e in query.edges:
matches = get_matches(e, d)
coord.append(matches)
if len(matches) == 1 and d[e.name] == None:
d[e.name] = e.domain.partitions[matches[0]]
for s in select:
mat = result[s.name]
agg = s.aggregate
var = s.value
expr = jx_expression_to_function(var)
val = expr(d)
if agg == "count":
if var == "." or var == None:
for c in itertools.product(*coord):
mat[c] += 1
continue
if val != None:
for c in itertools.product(*coord):
mat[c] += 1
else:
for c in itertools.product(*coord):
acc = mat[c]
if acc == None:
acc = windows.name2accumulator.get(agg)
if acc == None:
Log.error(
"select aggregate {{agg}} is not recognized",
agg=agg)
acc = acc(**s)
mat[c] = acc
acc.add(val)
for s in select:
if s.aggregate == "count":
continue
m = result[s.name]
for c, var in m.items():
if var != None:
m[c] = var.end()
from jx_python.containers.cube import Cube
return Cube(select, query.edges, result)
def es_terms_stats(esq, mvel, query):
select = listwrap(query.select)
facetEdges = [] # EDGES THAT WILL REQUIRE A FACET FOR EACH PART
termsEdges = FlatList()
specialEdge = None
special_index = -1
# A SPECIAL EDGE IS ONE THAT HAS AN UNDEFINED NUMBER OF PARTITIONS AT QUERY TIME
# FIND THE specialEdge, IF ONE
for f, tedge in enumerate(query.edges):
if tedge.domain.type in domains.KNOWN:
for p, part in enumerate(tedge.domain.partitions):
part.dataIndex = p
# FACETS ARE ONLY REQUIRED IF SQL JOIN ON DOMAIN IS REQUIRED (RANGE QUERY)
# OR IF WE ARE NOT SIMPLY COUNTING
# OR IF NO SCRIPTING IS ALLOWED (SOME OTHER CODE IS RESPONSIBLE FOR SETTING isFacet)
# OR IF WE JUST WANT TO FORCE IT :)
# OF COURSE THE default EDGE IS NOT EXPLICIT, SO MUST BE A TERM
facetEdges.append(tedge)
else:
if specialEdge:
Log.error("There is more than one open-ended edge: self can not be handled")
specialEdge = tedge
special_index = f
termsEdges.append(tedge)
if not specialEdge:
# WE SERIOUSLY WANT A SPECIAL EDGE, OTHERWISE WE WILL HAVE TOO MANY FACETS
# THE BIGGEST EDGE MAY BE COLLAPSED TO A TERM, MAYBE?
num_parts = 0
special_index = -1
for i, e in enumerate(facetEdges):
l = len(e.domain.partitions)
if ((e.value and is_variable_name(e.value)) or len(e.domain.dimension.fields) == 1) and l > num_parts:
num_parts = l
specialEdge = e
special_index = i
facetEdges.pop(special_index)
termsEdges.append(specialEdge)
total_facets = PRODUCT(len(f.domain.partitions) for f in facetEdges)*len(select)
if total_facets > 100:
# WE GOT A PROBLEM, LETS COUNT THE SIZE OF REALITY:
counts = esq.query({
"from": query.frum,
"select": {"aggregate": "count"},
"edges": facetEdges,
"where": query.where,
"limit": query.limit
})
esFacets = []
def add_facet(value, parts, cube):
if value:
esFacets.append(parts)
counts["count"].forall(add_facet)
Log.note("{{theory_count}} theoretical combinations, {{real_count}} actual combos found", real_count= len(esFacets), theory_count=total_facets)
if not esFacets:
# MAKE EMPTY CUBE
matricies = {}
dims = [len(e.domain.partitions) + (1 if e.allowNulls else 0) for e in query.edges]
for s in select:
matricies[s.name] = Matrix(*dims)
cube = Cube(query.select, query.edges, matricies)
cube.frum = query
return cube
else:
# GENERATE ALL COMBOS
esFacets = getAllEdges(facetEdges)
calcTerm = compileEdges2Term(mvel, termsEdges, FlatList())
term2parts = calcTerm.term2parts
if len(esFacets) * len(select) > 1000:
Log.error("not implemented yet") # WE HAVE SOME SERIOUS PERMUTATIONS, WE MUST ISSUE MULTIPLE QUERIES
pass
FromES = build_es_query(query)
for s in select:
for parts in esFacets:
condition = FlatList()
constants = FlatList()
name = [literal_field(s.name)]
for f, fedge in enumerate(facetEdges):
name.append(str(parts[f].dataIndex))
condition.append(buildCondition(mvel, fedge, parts[f]))
constants.append({"name": fedge.domain.name, "value": parts[f]})
condition.append(query.where)
name = ",".join(name)
FromES.facets[name] = {
"terms_stats": {
"key_field": calcTerm.field,
"value_field": s.value if is_variable_name(s.value) else None,
"value_script": mvel.compile_expression(s.value) if not is_variable_name(s.value) else None,
"size": coalesce(query.limit, 200000)
}
}
if condition:
FromES.facets[name].facet_filter = simplify_esfilter({"and": condition})
data = es_post(esq.es, FromES, query.limit)
if specialEdge.domain.type not in domains.KNOWN:
# WE BUILD THE PARTS BASED ON THE RESULTS WE RECEIVED
partitions = FlatList()
map = {}
for facetName, parts in data.facets.items():
for stats in parts.terms:
if not map[stats]:
part = {"value": stats, "name": stats}
partitions.append(part)
map[stats] = part
partitions.sort(specialEdge.domain.compare)
for p, part in enumerate(partitions):
part.dataIndex = p
specialEdge.domain.map = map
specialEdge.domain.partitions = partitions
# MAKE CUBE
matricies = {}
dims = [len(e.domain.partitions) + (1 if e.allowNulls else 0) for e in query.edges]
for s in select:
matricies[s.name] = Matrix(*dims)
name2agg = {s.name: aggregates[s.aggregate] for s in select}
# FILL CUBE
for edgeName, parts in data.facets.items():
temp = edgeName.split(",")
pre_coord = tuple(int(c) for c in temp[1:])
sname = temp[0]
for stats in parts.terms:
if specialEdge:
special = term2parts(stats.term)[0]
coord = pre_coord[:special_index]+(special.dataIndex, )+pre_coord[special_index:]
else:
coord = pre_coord
matricies[sname][coord] = stats[name2agg[sname]]
cube = Cube(query.select, query.edges, matricies)
cube.frum = query
return cube