def to_sql(self, schema, not_null=False, boolean=False):
default = self.default.to_sql(schema)
if len(self.terms) == 0:
return default
len_sep = LengthOp(self.separator).partial_eval()
no_sep = is_literal(len_sep) and len_sep.value == 0
sep = SQLang[self.separator].to_sql(schema)[0].sql.s
acc = []
for t in self.terms:
t = SQLang[t]
missing = t.missing().partial_eval()
term = t.to_sql(schema, not_null=True)[0].sql
if term.s:
term_sql = term.s
elif term.n:
term_sql = "cast(" + term.n + " as text)"
else:
term_sql = (SQL_CASE + SQL_WHEN + term.b + SQL_THEN +
quote_value("true") + SQL_ELSE +
quote_value("false") + SQL_END)
if no_sep:
sep_term = term_sql
else:
sep_term = sql_iso(sql_concat_text([sep, term_sql]))
if isinstance(missing, TrueOp):
acc.append(SQL_EMPTY_STRING)
elif missing:
acc.append(
SQL_CASE + SQL_WHEN +
sql_iso(missing.to_sql(schema, boolean=True)[0].sql.b) +
SQL_THEN + SQL_EMPTY_STRING + SQL_ELSE + sep_term +
SQL_END)
else:
acc.append(sep_term)
if no_sep:
expr_ = sql_concat_text(acc)
else:
expr_ = sql_call(
"SUBSTR", sql_concat_text(acc),
ConcatSQL(
LengthOp(self.separator).to_sql(schema)[0].sql.n, SQL_PLUS,
SQL_ONE))
return SQLScript(
expr=expr_,
data_type=STRING,
frum=self,
miss=self.missing(),
many=False,
schema=schema,
)
def to_sql(self, schema, not_null=False, boolean=False):
value = self.value
if value == None:
return wrap([{"name": "."}])
elif isinstance(value, text):
return wrap([{"name": ".", "sql": {"s": quote_value(value)}}])
elif is_number(value):
return wrap([{"name": ".", "sql": {"n": quote_value(value)}}])
elif value in [True, False]:
return wrap([{"name": ".", "sql": {"b": quote_value(value)}}])
else:
return wrap([{"name": ".", "sql": {"j": quote_value(self.json)}}])
def to_sql(self, schema, not_null=False, boolean=False):
term = SQLang[self.term].partial_eval()
if is_literal(term):
val = term.value
if isinstance(val, text):
sql = quote_value(len(val))
elif isinstance(val, (float, int)):
sql = quote_value(len(value2json(val)))
else:
return Null
else:
value = term.to_sql(schema, not_null=not_null)[0].sql.s
sql = ConcatSQL(SQL("LENGTH"), sql_iso(value))
return wrap([{"name": ".", "sql": {"n": sql}}])
def to_sql(self, schema, not_null=False, boolean=False):
prefix = SQLang[self.prefix].partial_eval()
if is_literal(prefix):
value = SQLang[self.value].partial_eval().to_sql(schema)[0].sql.s
prefix = prefix.value
if "%" in prefix or "_" in prefix:
for r in "\\_%":
prefix = prefix.replaceAll(r, "\\" + r)
sql = ConcatSQL(value, SQL_LIKE, quote_value(prefix + "%"),
SQL_ESCAPE, SQL("\\"))
else:
sql = ConcatSQL(value, SQL_LIKE, quote_value(prefix + "%"))
return wrap([{"name": ".", "sql": {"b": sql}}])
else:
return (SqlEqOp([SqlInstrOp([self.value, prefix]),
SQL_ONE]).partial_eval().to_sql())
def _insert(self, collection):
for nested_path, details in collection.items():
active_columns = wrap(list(details.active_columns))
rows = details.rows
num_rows = len(rows)
table_name = concat_field(self.name, nested_path)
if table_name == self.name:
# DO NOT REQUIRE PARENT OR ORDER COLUMNS
meta_columns = [GUID, UID]
else:
meta_columns = [UID, PARENT, ORDER]
all_columns = meta_columns + active_columns.es_column # ONLY THE PRIMITIVE VALUE COLUMNS
command = ConcatSQL(
SQL_INSERT,
quote_column(table_name),
sql_iso(sql_list(map(quote_column, all_columns))),
SQL_VALUES,
sql_list(
sql_iso(sql_list(quote_value(row.get(c)) for c in all_columns))
for row in unwrap(rows)
)
)
with self.db.transaction() as t:
t.execute(command)
def to_sql(self, schema, not_null=False, boolean=False):
pattern = quote_value(json2value(self.pattern.json))
value = self.var.to_sql(schema)[0].sql.s
return wrap([{
"name": ".",
"sql": {
"b": value + " REGEXP " + pattern
}
}])
def _edges_op(self, query, frum):
schema = frum
query = query.copy() # WE WILL BE MARKING UP THE QUERY
index_to_column = {} # MAP FROM INDEX TO COLUMN (OR SELECT CLAUSE)
outer_selects = [] # EVERY SELECT CLAUSE (NOT TO BE USED ON ALL TABLES, OF COURSE)
base_table, path = schema.snowflake.fact_name, schema.nested_path
nest_to_alias = {
nested_path: "__" + unichr(ord('a') + i) + "__"
for i, (nested_path, sub_table) in enumerate(self.snowflake.tables)
}
tables = []
for n, a in nest_to_alias.items():
if startswith_field(path, n):
tables.append({"nest": n, "alias": a})
tables = jx.sort(tables, {"value": {"length": "nest"}})
from_sql = sql_alias(quote_column(concat_field(base_table, tables[0].nest)), tables[0].alias)
for previous, t in zip(tables, tables[1::]):
from_sql += (
SQL_LEFT_JOIN + sql_alias(quote_column(concat_field(base_table, t.nest)), t.alias) +
SQL_ON + quote_column(t.alias, PARENT) + SQL_EQ + quote_column(previous.alias, UID)
)
main_filter = SQLang[query.where].to_sql(schema, boolean=True)[0].sql.b
# SHIFT THE COLUMN DEFINITIONS BASED ON THE NESTED QUERY DEPTH
ons = []
join_types = []
wheres = []
null_ons = [EXISTS_COLUMN + SQL_IS_NULL]
groupby = []
null_groupby = []
orderby = []
domains = []
select_clause = [SQL_ONE + EXISTS_COLUMN] + [quote_column(c.es_column) for c in self.snowflake.columns]
for edge_index, query_edge in enumerate(query.edges):
edge_alias = "e" + text(edge_index)
if query_edge.value:
edge_values = [p for c in SQLang[query_edge.value].to_sql(schema).sql for p in c.items()]
elif not query_edge.value and any(query_edge.domain.partitions.where):
case = SQL_CASE
for pp, p in enumerate(query_edge.domain.partitions):
w = SQLang[p.where].to_sql(schema)[0].sql.b
t = quote_value(pp)
case += SQL_WHEN + w + SQL_THEN + t
case += SQL_ELSE + SQL_NULL + SQL_END # quote value with length of partitions
edge_values = [("n", case)]
elif query_edge.range:
edge_values = SQLang[query_edge.range.min].to_sql(schema)[0].sql.items() + SQLang[query_edge.range.max].to_sql(schema)[
0].sql.items()
def _make_range_domain(self, domain, column_name):
width = (domain.max - domain.min) / domain.interval
digits = mo_math.floor(mo_math.log10(width - 1))
if digits == 0:
value = quote_column("a", "value")
else:
value = SQL_PLUS.join("1" + ("0" * j) + SQL_STAR +
text(chr(ord(b'a') + j)) + ".value"
for j in range(digits + 1))
if domain.interval == 1:
if domain.min == 0:
domain = (SQL_SELECT + sql_alias(value, column_name) +
SQL_FROM +
sql_alias(quote_column(DIGITS_TABLE), "a"))
else:
domain = (SQL_SELECT + sql_alias(
sql_iso(value) + SQL_PLUS + quote_value(domain.min),
column_name) + SQL_FROM +
sql_alias(quote_column(DIGITS_TABLE), "a"))
else:
if domain.min == 0:
domain = ConcatSQL(
SQL_SELECT,
sql_alias(value + SQL_STAR + quote_value(domain.interval),
column_name), SQL_FROM,
sql_alias(quote_column(DIGITS_TABLE), "a"))
else:
domain = ConcatSQL(
SQL_SELECT,
sql_alias(
sql_iso(value, SQL_STAR, quote_value(domain.interval))
+ SQL_PLUS + quote_value(domain.min), column_name),
SQL_FROM, sql_alias(quote_column(DIGITS_TABLE), "a"))
for j in range(digits):
domain += SQL_INNER_JOIN + sql_alias(
quote_column(DIGITS_TABLE), text(
chr(ord(b'a') + j + 1))) + SQL_ON + SQL_TRUE
domain += SQL_WHERE + value + " < " + quote_value(width)
return domain
def to_sql(self, schema, not_null=False, boolean=False):
test = SQLang[self.term].missing().to_sql(schema,
boolean=True)[0].sql.b
value = SQLang[self.term].to_sql(schema, not_null=True)[0].sql
acc = []
for t, v in value.items():
if t == "b":
acc.append(SQL_CASE + SQL_WHEN + sql_iso(test) + SQL_THEN +
SQL_NULL + SQL_WHEN + sql_iso(v) + SQL_THEN +
"'true'" + SQL_ELSE + "'false'" + SQL_END)
elif t == "s":
acc.append(v)
else:
acc.append("RTRIM(RTRIM(CAST" +
sql_iso(v + " as TEXT), " + quote_value("0")) +
", " + quote_value(".") + ")")
if not acc:
return wrap([{}])
elif len(acc) == 1:
return wrap([{"name": ".", "sql": {"s": acc[0]}}])
else:
return wrap([{"name": ".", "sql": {"s": sql_coalesce(acc)}}])
def request(self, method, path, headers):
now = Date.now()
self.inbound_rate.add(now)
ready = Signal(path)
# TEST CACHE
with self.cache_locker:
pair = self.cache.get(path)
if pair is None:
self.cache[path] = (ready, None, None, now)
if pair is not None:
# REQUEST IS IN THE QUEUE ALREADY, WAIT
ready, headers, response, then = pair
if response is None:
ready.wait()
with self.cache_locker:
ready, headers, response, timestamp = self.cache.get(path)
with self.db.transaction() as t:
t.execute("UPDATE cache SET timestamp=" + quote_value(now) +
" WHERE path=" + quote_value(path) +
" AND timestamp<" + quote_value(now))
return Response(response, status=200, headers=json.loads(headers))
# TEST DB
db_response = self.db.query(
"SELECT headers, response FROM cache WHERE path=" +
quote_value(path)).data
if db_response:
headers, response = db_response[0]
with self.db.transaction() as t:
t.execute("UPDATE cache SET timestamp=" + quote_value(now) +
" WHERE path=" + quote_value(path) +
" AND timestamp<" + quote_value(now))
with self.cache_locker:
self.cache[path] = (ready, headers, response.encode("latin1"),
now)
ready.go()
return Response(response, status=200, headers=json.loads(headers))
# MAKE A NETWORK REQUEST
self.todo.add((ready, method, path, headers, now))
ready.wait()
with self.cache_locker:
ready, headers, response, timestamp = self.cache[path]
return Response(response, status=200, headers=json.loads(headers))
push_child=push_child, # CAN NOT HANDLE TUPLES IN COLUMN
pull=pull,
sql=sql,
type=sql_type_to_json_type[sql_type],
column_alias=sql_name)
vals = [v for t, v in edge_values]
if query_edge.domain.type == "set":
domain_name = "d" + text(edge_index) + "c" + text(column_index)
domain_names = [domain_name]
if len(edge_names) > 1:
Log.error("Do not know how to handle")
if query_edge.value:
domain = SQL_UNION_ALL.join(
SQL_SELECT + sql_alias(
quote_value(coalesce(p.dataIndex, i)), "rownum") +
SQL_COMMA +
sql_alias(quote_value(p.value), domain_name)
for i, p in enumerate(query_edge.domain.partitions))
if query_edge.allowNulls:
domain += (SQL_UNION_ALL + SQL_SELECT + sql_alias(
quote_value(len(query_edge.domain.partitions)),
"rownum") + SQL_COMMA +
sql_alias(SQL_NULL, domain_name))
where = None
join_type = SQL_LEFT_JOIN if query_edge.allowNulls else SQL_INNER_JOIN
on_clause = (SQL_OR.join(
quote_column(edge_alias, k) + SQL_EQ + v
for k, v in zip(domain_names, vals)) + SQL_OR +
sql_iso(
quote_column(edge_alias, domain_name) +
def get_tuids(self, branch, revision, files):
"""
GET TUIDS FROM ENDPOINT, AND STORE IN DB
:param branch: BRANCH TO FIND THE REVISION/FILE
:param revision: THE REVISION NUNMBER
:param files: THE FULL PATHS TO THE FILES
:return: MAP FROM FILENAME TO TUID LIST
"""
# SCRUB INPUTS
revision = revision[:12]
files = [file.lstrip('/') for file in files]
with Timer(
"ask tuid service for {{num}} files at {{revision|left(12)}}",
{
"num": len(files),
"revision": revision
},
silent=not DEBUG or not self.enabled):
response = self.db.query(
"SELECT file, tuids FROM tuid WHERE revision=" +
quote_value(revision) + " AND file IN " + quote_list(files))
found = {file: json2value(tuids) for file, tuids in response.data}
try:
remaining = set(files) - set(found.keys())
new_response = None
if remaining:
request = wrap({
"from": "files",
"where": {
"and": [{
"eq": {
"revision": revision
}
}, {
"in": {
"path": remaining
}
}, {
"eq": {
"branch": branch
}
}]
},
"branch": branch,
"meta": {
"format": "list",
"request_time": Date.now()
}
})
if self.push_queue is not None:
if DEBUG:
Log.note(
"record tuid request to SQS: {{timestamp}}",
timestamp=request.meta.request_time)
self.push_queue.add(request)
else:
if DEBUG:
Log.note("no recorded tuid request")
if not self.enabled:
return found
new_response = http.post_json(self.endpoint,
json=request,
timeout=self.timeout)
if new_response.data and any(r.tuids
for r in new_response.data):
try:
with self.db.transaction() as transaction:
command = sql_insert(
"tuid", [{
"revision": revision,
"file": r.path,
"tuids": value2json(r.tuids)
} for r in new_response.data
if r.tuids != None])
transaction.execute(command)
except Exception as e:
Log.error("can not insert {{data|json}}",
data=new_response.data,
cause=e)
self.num_bad_requests = 0
found.update(
{r.path: r.tuids
for r in new_response.data} if new_response else {})
return found
except Exception as e:
self.num_bad_requests += 1
if self.enabled:
if "502 Bad Gateway" in e:
self.enabled = False
Log.alert(
"TUID service has problems (502 Bad Gateway)",
cause=e)
elif self.num_bad_requests >= MAX_BAD_REQUESTS:
self.enabled = False
Log.alert(
"TUID service has problems (given up trying to use it)",
cause=e)
else:
Log.alert("TUID service has problems.", cause=e)
Till(seconds=SLEEP_ON_ERROR).wait()
return found
def _set_op(self, query):
# GET LIST OF SELECTED COLUMNS
vars_ = UNION([
v.var for select in listwrap(query.select)
for v in select.value.vars()
])
schema = self.schema
known_vars = schema.keys()
active_columns = {".": set()}
for v in vars_:
for c in schema.leaves(v):
nest = c.nested_path[0]
active_columns.setdefault(nest, set()).add(c)
# ANY VARS MENTIONED WITH NO COLUMNS?
for v in vars_:
if not any(startswith_field(cname, v) for cname in known_vars):
active_columns["."].add(
Column(name=v,
jx_type=IS_NULL,
es_column=".",
es_index=".",
es_type='NULL',
nested_path=["."],
last_updated=Date.now()))
# EVERY COLUMN, AND THE INDEX IT TAKES UP
index_to_column = {} # MAP FROM INDEX TO COLUMN (OR SELECT CLAUSE)
index_to_uid = {} # FROM NESTED PATH TO THE INDEX OF UID
sql_selects = [
] # EVERY SELECT CLAUSE (NOT TO BE USED ON ALL TABLES, OF COURSE)
nest_to_alias = {
nested_path[0]: "__" + unichr(ord('a') + i) + "__"
for i, nested_path in enumerate(self.snowflake.query_paths)
}
sorts = []
if query.sort:
for select in query.sort:
col = SQLang[select.value].to_sql(schema)[0]
for t, sql in col.sql.items():
json_type = sql_type_to_json_type[t]
if json_type in STRUCT:
continue
column_number = len(sql_selects)
# SQL HAS ABS TABLE REFERENCE
column_alias = _make_column_name(column_number)
sql_selects.append(sql_alias(sql, column_alias))
if select.sort == -1:
sorts.append(quote_column(column_alias) + SQL_IS_NULL)
sorts.append(quote_column(column_alias) + " DESC")
else:
sorts.append(quote_column(column_alias) + SQL_IS_NULL)
sorts.append(quote_column(column_alias))
primary_doc_details = Data()
# EVERY SELECT STATEMENT THAT WILL BE REQUIRED, NO MATTER THE DEPTH
# WE WILL CREATE THEM ACCORDING TO THE DEPTH REQUIRED
nested_path = []
for step, sub_table in self.snowflake.tables:
nested_path.insert(0, step)
nested_doc_details = {
"sub_table": sub_table,
"children": [],
"index_to_column": {},
"nested_path": nested_path
}
# INSERT INTO TREE
if not primary_doc_details:
primary_doc_details = nested_doc_details
else:
def place(parent_doc_details):
if startswith_field(step,
parent_doc_details['nested_path'][0]):
for c in parent_doc_details['children']:
if place(c):
return True
parent_doc_details['children'].append(
nested_doc_details)
place(primary_doc_details)
alias = nested_doc_details['alias'] = nest_to_alias[step]
# WE ALWAYS ADD THE UID
column_number = index_to_uid[step] = nested_doc_details[
'id_coord'] = len(sql_selects)
sql_select = quote_column(alias, UID)
sql_selects.append(
sql_alias(sql_select, _make_column_name(column_number)))
if step != ".":
# ID AND ORDER FOR CHILD TABLES
index_to_column[column_number] = ColumnMapping(
sql=sql_select,
type="number",
nested_path=nested_path,
column_alias=_make_column_name(column_number))
column_number = len(sql_selects)
sql_select = quote_column(alias, ORDER)
sql_selects.append(
sql_alias(sql_select, _make_column_name(column_number)))
index_to_column[column_number] = ColumnMapping(
sql=sql_select,
type="number",
nested_path=nested_path,
column_alias=_make_column_name(column_number))
# WE DO NOT NEED DATA FROM TABLES WE REQUEST NOTHING FROM
if step not in active_columns:
continue
# ADD SQL SELECT COLUMNS FOR EACH jx SELECT CLAUSE
si = 0
for select in listwrap(query.select):
try:
column_number = len(sql_selects)
select.pull = get_column(column_number)
db_columns = SQLang[select.value].partial_eval().to_sql(
schema)
for column in db_columns:
for t, unsorted_sql in column.sql.items():
json_type = sql_type_to_json_type[t]
if json_type in STRUCT:
continue
column_number = len(sql_selects)
column_alias = _make_column_name(column_number)
sql_selects.append(
sql_alias(unsorted_sql, column_alias))
if startswith_field(schema.path, step) and is_op(
select.value, LeavesOp):
# ONLY FLATTEN primary_nested_path AND PARENTS, NOT CHILDREN
index_to_column[
column_number] = nested_doc_details[
'index_to_column'][
column_number] = ColumnMapping(
push_name=literal_field(
get_property_name(
concat_field(
select.name,
column.name))),
push_child=".",
push_column_name=
get_property_name(
concat_field(
select.name,
column.name)),
push_column=si,
pull=get_column(column_number),
sql=unsorted_sql,
type=json_type,
column_alias=column_alias,
nested_path=nested_path)
si += 1
else:
index_to_column[
column_number] = nested_doc_details[
'index_to_column'][
column_number] = ColumnMapping(
push_name=select.name,
push_child=column.name,
push_column_name=select.name,
push_column=si,
pull=get_column(column_number),
sql=unsorted_sql,
type=json_type,
column_alias=column_alias,
nested_path=nested_path)
finally:
si += 1
where_clause = BooleanOp(query.where).partial_eval().to_sql(
schema, boolean=True)[0].sql.b
unsorted_sql = self._make_sql_for_one_nest_in_set_op(
".", sql_selects, where_clause, active_columns, index_to_column)
for n, _ in self.snowflake.tables:
sorts.append(quote_column(COLUMN + text(index_to_uid[n])))
ordered_sql = ConcatSQL(SQL_SELECT, SQL_STAR, SQL_FROM,
sql_iso(unsorted_sql), SQL_ORDERBY,
sql_list(sorts), SQL_LIMIT,
quote_value(query.limit))
result = self.db.query(ordered_sql)
def _accumulate_nested(rows, row, nested_doc_details, parent_doc_id,
parent_id_coord):
"""
:param rows: REVERSED STACK OF ROWS (WITH push() AND pop())
:param row: CURRENT ROW BEING EXTRACTED
:param nested_doc_details: {
"nested_path": wrap_nested_path(nested_path),
"index_to_column": map from column number to column details
"children": all possible direct decedents' nested_doc_details
}
:param parent_doc_id: the id of the parent doc (for detecting when to step out of loop)
:param parent_id_coord: the column number for the parent id (so we ca extract from each row)
:return: the nested property (usually an array)
"""
previous_doc_id = None
doc = Null
output = []
id_coord = nested_doc_details['id_coord']
while True:
doc_id = row[id_coord]
if doc_id == None or (parent_id_coord is not None and
row[parent_id_coord] != parent_doc_id):
rows.append(
row
) # UNDO PREVIOUS POP (RECORD IS NOT A NESTED RECORD OF parent_doc)
return output
if doc_id != previous_doc_id:
previous_doc_id = doc_id
doc = Null
curr_nested_path = nested_doc_details['nested_path'][0]
index_to_column = nested_doc_details[
'index_to_column'].items()
for i, c in index_to_column:
value = row[i]
if is_list(query.select) or is_op(
query.select.value, LeavesOp):
# ASSIGN INNER PROPERTIES
relative_field = concat_field(
c.push_name, c.push_child)
else: # FACT IS EXPECTED TO BE A SINGLE VALUE, NOT AN OBJECT
relative_field = c.push_child
if relative_field == ".":
if exists(value):
doc = value
elif exists(value):
if doc is Null:
doc = Data()
doc[relative_field] = value
for child_details in nested_doc_details['children']:
# EACH NESTED TABLE MUST BE ASSEMBLED INTO A LIST OF OBJECTS
child_id = row[child_details['id_coord']]
if child_id is not None:
nested_value = _accumulate_nested(
rows, row, child_details, doc_id, id_coord)
if nested_value != None:
push_name = child_details['nested_path'][0]
if is_list(query.select) or is_op(
query.select.value, LeavesOp):
# ASSIGN INNER PROPERTIES
relative_field = relative_field(
push_name, curr_nested_path)
else: # FACT IS EXPECTED TO BE A SINGLE VALUE, NOT AN OBJECT
relative_field = "."
if relative_field == ".":
doc = unwraplist(nested_value)
else:
doc[relative_field] = unwraplist(nested_value)
output.append(doc)
try:
row = rows.pop()
except IndexError:
return output
cols = tuple(
[i for i in index_to_column.values() if i.push_name != None])
rows = list(reversed(unwrap(result.data)))
if rows:
row = rows.pop()
data = _accumulate_nested(rows, row, primary_doc_details, None,
None)
else:
data = result.data
if query.format == "cube":
# for f, full_name in self.snowflake.tables:
# if f != '.' or (test_dots(cols) and is_list(query.select)):
# num_rows = len(result.data)
# num_cols = MAX([c.push_column for c in cols]) + 1 if len(cols) else 0
# map_index_to_name = {c.push_column: c.push_column_name for c in cols}
# temp_data = [[None] * num_rows for _ in range(num_cols)]
# for rownum, d in enumerate(result.data):
# for c in cols:
# if c.push_child == ".":
# temp_data[c.push_column][rownum] = c.pull(d)
# else:
# column = temp_data[c.push_column][rownum]
# if column is None:
# column = temp_data[c.push_column][rownum] = {}
# column[c.push_child] = c.pull(d)
# output = Data(
# meta={"format": "cube"},
# data={n: temp_data[c] for c, n in map_index_to_name.items()},
# edges=[{
# "name": "rownum",
# "domain": {
# "type": "rownum",
# "min": 0,
# "max": num_rows,
# "interval": 1
# }
# }]
# )
# return output
if is_list(query.select) or is_op(query.select.value, LeavesOp):
num_rows = len(data)
temp_data = {
c.push_column_name: [None] * num_rows
for c in cols
}
for rownum, d in enumerate(data):
for c in cols:
temp_data[c.push_column_name][rownum] = d[c.push_name]
return Data(meta={"format": "cube"},
data=temp_data,
edges=[{
"name": "rownum",
"domain": {
"type": "rownum",
"min": 0,
"max": num_rows,
"interval": 1
}
}])
else:
num_rows = len(data)
map_index_to_name = {
c.push_column: c.push_column_name
for c in cols
}
temp_data = [data]
return Data(meta={"format": "cube"},
data={
n: temp_data[c]
for c, n in map_index_to_name.items()
},
edges=[{
"name": "rownum",
"domain": {
"type": "rownum",
"min": 0,
"max": num_rows,
"interval": 1
}
}])
elif query.format == "table":
# for f, _ in self.snowflake.tables:
# if frum.endswith(f):
# num_column = MAX([c.push_column for c in cols]) + 1
# header = [None] * num_column
# for c in cols:
# header[c.push_column] = c.push_column_name
#
# output_data = []
# for d in result.data:
# row = [None] * num_column
# for c in cols:
# set_column(row, c.push_column, c.push_child, c.pull(d))
# output_data.append(row)
#
# return Data(
# meta={"format": "table"},
# header=header,
# data=output_data
# )
if is_list(query.select) or is_op(query.select.value, LeavesOp):
column_names = [None] * (max(c.push_column for c in cols) + 1)
for c in cols:
column_names[c.push_column] = c.push_column_name
temp_data = []
for rownum, d in enumerate(data):
row = [None] * len(column_names)
for c in cols:
row[c.push_column] = d[c.push_name]
temp_data.append(row)
return Data(meta={"format": "table"},
header=column_names,
data=temp_data)
else:
column_names = listwrap(query.select).name
return Data(meta={"format": "table"},
header=column_names,
data=[[d] for d in data])
else:
# for f, _ in self.snowflake.tables:
# if frum.endswith(f) or (test_dots(cols) and is_list(query.select)):
# data = []
# for d in result.data:
# row = Data()
# for c in cols:
# if c.push_child == ".":
# row[c.push_name] = c.pull(d)
# elif c.num_push_columns:
# tuple_value = row[c.push_name]
# if not tuple_value:
# tuple_value = row[c.push_name] = [None] * c.num_push_columns
# tuple_value[c.push_child] = c.pull(d)
# else:
# row[c.push_name][c.push_child] = c.pull(d)
#
# data.append(row)
#
# return Data(
# meta={"format": "list"},
# data=data
# )
if is_list(query.select) or is_op(query.select.value, LeavesOp):
temp_data = []
for rownum, d in enumerate(data):
row = {}
for c in cols:
row[c.push_column_name] = d[c.push_name]
temp_data.append(row)
return Data(meta={"format": "list"}, data=temp_data)
else:
return Data(meta={"format": "list"}, data=data)
def _groupby_op(self, query, schema):
base_table = schema.snowflake.fact_name
path = schema.nested_path
# base_table, path = tail_field(frum)
# schema = self.snowflake.tables[path].schema
index_to_column = {}
nest_to_alias = {
nested_path: "__" + unichr(ord('a') + i) + "__"
for i, nested_path in enumerate(self.schema.snowflake.query_paths)
}
tables = []
for n, a in nest_to_alias.items():
if startswith_field(path, n):
tables.append({"nest": n, "alias": a})
tables = jx.sort(tables, {"value": {"length": "nest"}})
from_sql = join_field(
[base_table] + split_field(tables[0].nest)) + " " + tables[0].alias
previous = tables[0]
for t in tables[1::]:
from_sql += (SQL_LEFT_JOIN +
quote_column(concat_field(base_table, t.nest)) + " " +
t.alias + SQL_ON + quote_column(t.alias, PARENT) +
SQL_EQ + quote_column(previous.alias, UID))
selects = []
groupby = []
for i, e in enumerate(query.groupby):
for edge_sql in SQLang[e.value].to_sql(schema):
column_number = len(selects)
sql_type, sql = edge_sql.sql.items()[0]
if sql is SQL_NULL and not e.value.var in schema.keys():
Log.error("No such column {{var}}", var=e.value.var)
column_alias = _make_column_name(column_number)
groupby.append(sql)
selects.append(sql_alias(sql, column_alias))
if edge_sql.nested_path == ".":
select_name = edge_sql.name
else:
select_name = "."
index_to_column[column_number] = ColumnMapping(
is_edge=True,
push_name=e.name,
push_column_name=e.name.replace("\\.", "."),
push_column=i,
push_child=select_name,
pull=get_column(column_number),
sql=sql,
column_alias=column_alias,
type=sql_type_to_json_type[sql_type])
for i, select in enumerate(listwrap(query.select)):
column_number = len(selects)
sql_type, sql = SQLang[select.value].to_sql(
schema)[0].sql.items()[0]
if sql == 'NULL' and not select.value.var in schema.keys():
Log.error("No such column {{var}}", var=select.value.var)
# AGGREGATE
if select.value == "." and select.aggregate == "count":
sql = sql_count(SQL_ONE)
else:
sql = sql_call(sql_aggs[select.aggregate], sql)
if select.default != None:
sql = sql_coalesce([sql, quote_value(select.default)])
selects.append(sql_alias(sql, select.name))
index_to_column[column_number] = ColumnMapping(
push_name=select.name,
push_column_name=select.name,
push_column=i + len(query.groupby),
push_child=".",
pull=get_column(column_number),
sql=sql,
column_alias=quote_column(select.name),
type=sql_type_to_json_type[sql_type])
for w in query.window:
selects.append(self._window_op(self, query, w))
where = SQLang[query.where].to_sql(schema)[0].sql.b
command = (SQL_SELECT + (sql_list(selects)) + SQL_FROM + from_sql +
SQL_WHERE + where + SQL_GROUPBY + sql_list(groupby))
if query.sort:
command += SQL_ORDERBY + sql_list(
sql_iso(sql[t]) + SQL_IS_NULL + "," + sql[t] +
(" DESC" if s.sort == -1 else "")
for s, sql in [(s, SQLang[s.value].to_sql(schema)[0].sql)
for s in query.sort] for t in "bns" if sql[t])
return command, index_to_column
def update(self, command):
"""
:param command: EXPECTING dict WITH {"set": s, "clear": c, "where": w} FORMAT
"""
command = wrap(command)
clear_columns = set(listwrap(command['clear']))
# REJECT DEEP UPDATES
touched_columns = command.set.keys() | clear_columns
for c in self.schema.columns:
if c.name in touched_columns and len(c.nested_path) > 1:
Log.error("Deep update not supported")
# ADD NEW COLUMNS
where = jx_expression(command.where) or TRUE
_vars = where.vars()
_map = {
v: c.es_column
for v in _vars for c in self.columns.get(v, Null)
if c.jx_type not in STRUCT
}
where_sql = where.map(_map).to_sql(self.schema)[0].sql.b
new_columns = set(command.set.keys()) - set(
c.name for c in self.schema.columns)
for new_column_name in new_columns:
nested_value = command.set[new_column_name]
ctype = get_jx_type(nested_value)
column = Column(name=new_column_name,
jx_type=ctype,
es_index=self.name,
es_type=json_type_to_sqlite_type(ctype),
es_column=typed_column(new_column_name, ctype),
last_updated=Date.now())
self.add_column(column)
# UPDATE THE NESTED VALUES
for nested_column_name, nested_value in command.set.items():
if get_jx_type(nested_value) == "nested":
nested_table_name = concat_field(self.name, nested_column_name)
nested_table = nested_tables[nested_column_name]
self_primary_key = sql_list(
quote_column(c.es_column) for u in self.uid
for c in self.columns[u])
extra_key_name = UID + text(len(self.uid))
extra_key = [e
for e in nested_table.columns[extra_key_name]][0]
sql_command = (
SQL_DELETE + SQL_FROM + quote_column(nested_table.name) +
SQL_WHERE + "EXISTS" +
sql_iso(SQL_SELECT + SQL_ONE + SQL_FROM +
sql_alias(quote_column(nested_table.name), "n") +
SQL_INNER_JOIN +
sql_iso(SQL_SELECT + self_primary_key + SQL_FROM +
quote_column(abs_schema.fact) + SQL_WHERE +
where_sql) + " t ON " +
SQL_AND.join(
quote_column("t", c.es_column) + SQL_EQ +
quote_column("n", c.es_column)
for u in self.uid for c in self.columns[u])))
self.db.execute(sql_command)
# INSERT NEW RECORDS
if not nested_value:
continue
doc_collection = {}
for d in listwrap(nested_value):
nested_table.flatten(d,
Data(),
doc_collection,
path=nested_column_name)
prefix = SQL_INSERT + quote_column(nested_table.name) + sql_iso(
sql_list([self_primary_key] + [quote_column(extra_key)] + [
quote_column(c.es_column)
for c in doc_collection.get(".", Null).active_columns
]))
# BUILD THE PARENT TABLES
parent = (SQL_SELECT + self_primary_key + SQL_FROM +
quote_column(abs_schema.fact) + SQL_WHERE +
jx_expression(command.where).to_sql(schema))
# BUILD THE RECORDS
children = SQL_UNION_ALL.join(
SQL_SELECT +
sql_alias(quote_value(i), extra_key.es_column) +
SQL_COMMA + sql_list(
sql_alias(quote_value(row[c.name]),
quote_column(c.es_column))
for c in doc_collection.get(".", Null).active_columns)
for i, row in enumerate(
doc_collection.get(".", Null).rows))
sql_command = (prefix + SQL_SELECT + sql_list([
quote_column("p", c.es_column) for u in self.uid
for c in self.columns[u]
] + [quote_column("c", extra_key)] + [
quote_column("c", c.es_column)
for c in doc_collection.get(".", Null).active_columns
]) + SQL_FROM + sql_iso(parent) + " p" + SQL_INNER_JOIN +
sql_iso(children) + " c" + SQL_ON + SQL_TRUE)
self.db.execute(sql_command)
# THE CHILD COLUMNS COULD HAVE EXPANDED
# ADD COLUMNS TO SELF
for n, cs in nested_table.columns.items():
for c in cs:
column = Column(name=c.name,
jx_type=c.jx_type,
es_type=c.es_type,
es_index=c.es_index,
es_column=c.es_column,
nested_path=[nested_column_name] +
c.nested_path,
last_updated=Date.now())
if c.name not in self.columns:
self.columns[column.name] = {column}
elif c.jx_type not in [
c.jx_type for c in self.columns[c.name]
]:
self.columns[column.name].add(column)
command = ConcatSQL(
SQL_UPDATE, quote_column(self.name), SQL_SET,
sql_list([
quote_column(c.es_column) + SQL_EQ +
quote_value(get_if_type(v, c.jx_type))
for c in self.schema.columns
if c.jx_type != NESTED and len(c.nested_path) == 1
for v in [command.set[c.name]] if v != None
] + [
quote_column(c.es_column) + SQL_EQ + SQL_NULL
for c in self.schema.columns
if (c.name in clear_columns and command.set[c.name] != None
and c.jx_type != NESTED and len(c.nested_path) == 1)
]), SQL_WHERE, where_sql)
with self.db.transaction() as t:
t.execute(command)
def to_sql(self, schema, not_null=False, boolean=False):
return wrap([{"name": ".", "sql": {"n": quote_value(self.value)}}])
pull=pull,
sql=sql,
type=sql_type_to_json_type[sql_type],
column_alias=sql_name
)
vals = [v for t, v in edge_values]
if query_edge.domain.type == "set":
domain_name = "d" + text(edge_index) + "c" + text(column_index)
domain_names = [domain_name]
if len(edge_names) > 1:
Log.error("Do not know how to handle")
if query_edge.value:
domain = SQL_UNION_ALL.join(
SQL_SELECT +
sql_alias(quote_value(coalesce(p.dataIndex, i)), "rownum") + SQL_COMMA +
sql_alias(quote_value(p.value), domain_name)
for i, p in enumerate(query_edge.domain.partitions)
)
if query_edge.allowNulls:
domain += (
SQL_UNION_ALL + SQL_SELECT +
sql_alias(quote_value(len(query_edge.domain.partitions)), "rownum") + SQL_COMMA +
sql_alias(SQL_NULL, domain_name)
)
where = None
join_type = SQL_LEFT_JOIN if query_edge.allowNulls else SQL_INNER_JOIN
on_clause = (
SQL_OR.join(
quote_column(edge_alias, k) + SQL_EQ + v
for k, v in zip(domain_names, vals)
