def __init__(self, host, index, alias=None, name=None, port=9200, settings=None):
global _elasticsearch
if hasattr(self, "settings"):
return
from pyLibrary.queries.containers.lists import ListContainer
from pyLibrary.env import elasticsearch as _elasticsearch
self.settings = settings
self.default_name = coalesce(name, alias, index)
self.default_es = _elasticsearch.Cluster(settings=settings)
self.todo = Queue("refresh metadata", max=100000, unique=True)
table_columns = metadata_tables()
column_columns = metadata_columns()
self.tables = ListContainer("meta.tables", [], wrap({c.name: c for c in table_columns}))
self.columns = ListContainer("meta.columns", [], wrap({c.name: c for c in column_columns}))
self.columns.insert(column_columns)
self.columns.insert(table_columns)
# TODO: fix monitor so it does not bring down ES
if ENABLE_META_SCAN:
self.worker = Thread.run("refresh metadata", self.monitor)
else:
self.worker = Thread.run("refresh metadata", self.not_monitor)
return
def query(self, q):
frum = self
if is_aggs(q):
return cube_aggs(frum, q)
columns = wrap({s.name: s for s in self.select + self.edges})
# DEFER TO ListContainer
from pyLibrary.queries.containers.lists import ListContainer
frum = ListContainer(name="", data=frum.values(), schema=columns)
return frum.query(q)
def query(self, q):
frum = self
if is_aggs(q):
return cube_aggs(frum, q)
columns = wrap({s.name: s for s in self.select + self.edges})
# DEFER TO ListContainer
from pyLibrary.queries.containers.lists import ListContainer
frum = ListContainer(name="", data=frum.values(), schema=columns)
return frum.query(q)
def __init__(self,
host,
index,
alias=None,
name=None,
port=9200,
settings=None):
global _elasticsearch
if hasattr(self, "settings"):
return
from pyLibrary.queries.containers.lists import ListContainer
from pyLibrary.env import elasticsearch as _elasticsearch
self.settings = settings
self.default_name = coalesce(name, alias, index)
self.default_es = _elasticsearch.Cluster(settings=settings)
self.todo = Queue("refresh metadata", max=100000, unique=True)
self.meta = Dict()
table_columns = metadata_tables()
column_columns = metadata_columns()
self.meta.tables = ListContainer(
"meta.tables", [], wrap({c.name: c
for c in table_columns}))
self.meta.columns = ListContainer(
"meta.columns", [], wrap({c.name: c
for c in column_columns}))
self.meta.columns.insert(column_columns)
self.meta.columns.insert(table_columns)
# TODO: fix monitor so it does not bring down ES
if ENABLE_META_SCAN:
self.worker = Thread.run("refresh metadata", self.monitor)
else:
self.worker = Thread.run("refresh metadata", self.not_monitor)
return
class FromESMetadata(object):
"""
QUERY THE METADATA
"""
def __new__(cls, *args, **kwargs):
global singlton
if singlton:
return singlton
else:
singlton = object.__new__(cls)
return singlton
@use_settings
def __init__(self, host, index, alias=None, name=None, port=9200, settings=None):
global _elasticsearch
if hasattr(self, "settings"):
return
from pyLibrary.queries.containers.lists import ListContainer
from pyLibrary.env import elasticsearch as _elasticsearch
self.settings = settings
self.default_name = coalesce(name, alias, index)
self.default_es = _elasticsearch.Cluster(settings=settings)
self.todo = Queue("refresh metadata", max=100000, unique=True)
table_columns = metadata_tables()
column_columns = metadata_columns()
self.tables = ListContainer("meta.tables", [], wrap({c.name: c for c in table_columns}))
self.columns = ListContainer("meta.columns", [], wrap({c.name: c for c in column_columns}))
self.columns.insert(column_columns)
self.columns.insert(table_columns)
# TODO: fix monitor so it does not bring down ES
if ENABLE_META_SCAN:
self.worker = Thread.run("refresh metadata", self.monitor)
else:
self.worker = Thread.run("refresh metadata", self.not_monitor)
return
@property
def query_path(self):
return None
@property
def url(self):
return self.default_es.path + "/" + self.default_name.replace(".", "/")
def get_table(self, table_name):
with self.tables.locker:
return self.tables.query({"where": {"eq": {"name": table_name}}})
def upsert_column(self, c):
existing_columns = filter(lambda r: r.table == c.table and r.abs_name == c.abs_name, self.columns.data)
if not existing_columns:
self.columns.add(c)
cols = filter(lambda r: r.table == "meta.columns", self.columns.data)
for cc in cols:
cc.partitions = cc.cardinality = cc.last_updated = None
self.todo.add(c)
self.todo.extend(cols)
else:
set_default(existing_columns[0], c)
self.todo.add(existing_columns[0])
# TEST CONSISTENCY
for c, d in product(list(self.todo.queue), list(self.todo.queue)):
if c.abs_name==d.abs_name and c.table==d.table and c!=d:
Log.error("")
def _get_columns(self, table=None):
# TODO: HANDLE MORE THEN ONE ES, MAP TABLE SHORT_NAME TO ES INSTANCE
alias_done = set()
index = split_field(table)[0]
query_path = split_field(table)[1:]
metadata = self.default_es.get_metadata(index=index)
for index, meta in qb.sort(metadata.indices.items(), {"value": 0, "sort": -1}):
for _, properties in meta.mappings.items():
columns = _elasticsearch.parse_properties(index, None, properties.properties)
columns = columns.filter(lambda r: not r.abs_name.startswith("other.") and not r.abs_name.startswith("previous_values.cf_")) # TODO: REMOVE WHEN jobs PROPERTY EXPLOSION IS CONTAINED
with Timer("upserting {{num}} columns", {"num": len(columns)}, debug=DEBUG):
with self.columns.locker:
for c in columns:
# ABSOLUTE
c.table = join_field([index]+query_path)
self.upsert_column(c)
for alias in meta.aliases:
# ONLY THE LATEST ALIAS IS CHOSEN TO GET COLUMNS
if alias in alias_done:
continue
alias_done.add(alias)
c = copy(c)
c.table = join_field([alias]+query_path)
self.upsert_column(c)
def query(self, _query):
return self.columns.query(Query(set_default(
{
"from": self.columns,
"sort": ["table", "name"]
},
_query.as_dict()
)))
def get_columns(self, table):
"""
RETURN METADATA COLUMNS
"""
with self.columns.locker:
columns = qb.sort(filter(lambda r: r.table == table, self.columns.data), "name")
if columns:
return columns
self._get_columns(table=table)
with self.columns.locker:
columns = qb.sort(filter(lambda r: r.table == table, self.columns.data), "name")
if columns:
return columns
# self._get_columns(table=table)
Log.error("no columns for {{table}}", table=table)
def _update_cardinality(self, c):
"""
QUERY ES TO FIND CARDINALITY AND PARTITIONS FOR A SIMPLE COLUMN
"""
if c.type in ["object", "nested"]:
Log.error("not supported")
try:
if c.table == "meta.columns":
with self.columns.locker:
partitions = qb.sort([g[c.abs_name] for g, _ in qb.groupby(self.columns, c.abs_name) if g[c.abs_name] != None])
self.columns.update({
"set": {
"partitions": partitions,
"count": len(self.columns),
"cardinality": len(partitions),
"last_updated": Date.now()
},
"where": {"eq": {"table": c.table, "abs_name": c.abs_name}}
})
return
if c.table == "meta.tables":
with self.columns.locker:
partitions = qb.sort([g[c.abs_name] for g, _ in qb.groupby(self.tables, c.abs_name) if g[c.abs_name] != None])
self.columns.update({
"set": {
"partitions": partitions,
"count": len(self.tables),
"cardinality": len(partitions),
"last_updated": Date.now()
},
"where": {"eq": {"table": c.table, "name": c.name}}
})
return
es_index = c.table.split(".")[0]
result = self.default_es.post("/"+es_index+"/_search", data={
"aggs": {c.name: _counting_query(c)},
"size": 0
})
r = result.aggregations.values()[0]
count = result.hits.total
cardinality = coalesce(r.value, r._nested.value)
if cardinality == None:
Log.error("logic error")
query = Dict(size=0)
if c.type in ["object", "nested"]:
Log.note("{{field}} has {{num}} parts", field=c.name, num=cardinality)
with self.columns.locker:
self.columns.update({
"set": {
"count": count,
"cardinality": cardinality,
"last_updated": Date.now()
},
"clear": ["partitions"],
"where": {"eq": {"table": c.table, "name": c.name}}
})
return
elif cardinality > 1000 or (count >= 30 and cardinality == count) or (count >= 1000 and cardinality / count > 0.99):
Log.note("{{field}} has {{num}} parts", field=c.name, num=cardinality)
with self.columns.locker:
self.columns.update({
"set": {
"count": count,
"cardinality": cardinality,
"last_updated": Date.now()
},
"clear": ["partitions"],
"where": {"eq": {"table": c.table, "name": c.name}}
})
return
elif c.type in _elasticsearch.ES_NUMERIC_TYPES and cardinality > 30:
Log.note("{{field}} has {{num}} parts", field=c.name, num=cardinality)
with self.columns.locker:
self.columns.update({
"set": {
"count": count,
"cardinality": cardinality,
"last_updated": Date.now()
},
"clear": ["partitions"],
"where": {"eq": {"table": c.table, "name": c.name}}
})
return
elif c.nested_path:
query.aggs[literal_field(c.name)] = {
"nested": {"path": listwrap(c.nested_path)[0]},
"aggs": {"_nested": {"terms": {"field": c.abs_name, "size": 0}}}
}
else:
query.aggs[literal_field(c.name)] = {"terms": {"field": c.abs_name, "size": 0}}
result = self.default_es.post("/"+es_index+"/_search", data=query)
aggs = result.aggregations.values()[0]
if aggs._nested:
parts = qb.sort(aggs._nested.buckets.key)
else:
parts = qb.sort(aggs.buckets.key)
Log.note("{{field}} has {{parts}}", field=c.name, parts=parts)
with self.columns.locker:
self.columns.update({
"set": {
"count": count,
"cardinality": cardinality,
"partitions": parts,
"last_updated": Date.now()
},
"where": {"eq": {"table": c.table, "abs_name": c.abs_name}}
})
except Exception, e:
if "IndexMissingException" in e and c.table.startswith("testing"):
Log.alert("{{col.table}} does not exist", col=c)
else:
self.columns.update({
"set": {
"last_updated": Date.now()
},
"clear":[
"count",
"cardinality",
"partitions",
],
"where": {"eq": {"table": c.table, "abs_name": c.abs_name}}
})
Log.warning("Could not get {{col.table}}.{{col.abs_name}} info", col=c, cause=e)
