def denormalized(self):
"""
THE INTERNAL STRUCTURE FOR THE COLUMN METADATA IS VERY DIFFERENT FROM
THE DENORMALIZED PERSPECITVE. THIS PROVIDES THAT PERSPECTIVE FOR QUERIES
"""
with self.locker:
self._update_meta()
output = [
{
"table": c.es_index,
"name": untype_path(c.name),
"cardinality": c.cardinality,
"es_column": c.es_column,
"es_index": c.es_index,
"last_updated": c.last_updated,
"count": c.count,
"nested_path": [unnest_path(n) for n in c.nested_path],
"es_type": c.es_type,
"type": c.jx_type,
} for tname, css in self.data.items()
for cname, cs in css.items() for c in cs
if c.jx_type not in STRUCT # and c.es_column != "_id"
]
from jx_python.containers.list_usingPythonList import ListContainer
return ListContainer(
self.name,
data=output,
schema=jx_base.Schema(META_COLUMNS_NAME, SIMPLE_METADATA_COLUMNS),
)
def __init__(self,
host,
index,
sql_file='metadata.sqlite',
alias=None,
name=None,
port=9200,
kwargs=None):
if hasattr(self, "settings"):
return
self.too_old = TOO_OLD
self.settings = kwargs
self.default_name = coalesce(name, alias, index)
self.es_cluster = elasticsearch.Cluster(kwargs=kwargs)
self.index_does_not_exist = set()
self.todo = Queue("refresh metadata", max=100000, unique=True)
self.index_to_alias = Relation_usingList()
self.es_metadata = Null
self.metadata_last_updated = Date.now() - OLD_METADATA
self.meta = Data()
self.meta.columns = ColumnList()
self.alias_to_query_paths = {
"meta.columns": [['.']],
"meta.tables": [['.']]
}
self.alias_last_updated = {
"meta.columns": Date.now(),
"meta.tables": Date.now()
}
table_columns = metadata_tables()
self.meta.tables = ListContainer(
"meta.tables",
[
# TableDesc("meta.columns", None, ".", Date.now()),
# TableDesc("meta.tables", None, ".", Date.now())
],
jx_base.Schema(".", table_columns))
self.meta.columns.extend(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 __init__(self, host, index, alias=None, name=None, port=9200, kwargs=None):
if hasattr(self, "settings"):
return
self.settings = kwargs
self.too_old = TOO_OLD
self.es_cluster = elasticsearch.Cluster(kwargs=kwargs)
self.index_does_not_exist = set()
self.todo = Queue("refresh metadata", max=100000, unique=True)
self.meta = Data()
self.meta.columns = ColumnList(self.es_cluster)
self.meta.columns.extend(META_TABLES_DESC.columns)
self.meta.tables = ListContainer(META_TABLES_NAME, [], jx_base.Schema(".", META_TABLES_DESC.columns))
self.meta.table.extend([META_COLUMNS_DESC, META_TABLES_DESC])
self.alias_to_query_paths = {}
for i, settings in self.es_cluster.get_metadata().indices.items():
if len(settings.aliases) == 0:
alias = i
elif len(settings.aliases) == 1:
alias = first(settings.aliases)
else:
Log.error("expecting only one alias per index")
desc = TableDesc(
name=alias,
url=None,
query_path=ROOT_PATH,
last_updated=Date.MIN,
columns=[]
)
self.meta.tables.add(desc)
self.alias_to_query_paths[alias] = [desc.query_path]
self.alias_to_query_paths[self._find_alias(alias)] = [desc.query_path]
# WE MUST PAUSE?
# 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("not refresh metadata", self.not_monitor)
return
