def graphql_match(query, template):
query = graphql.parse(query)
query = graphql.print_ast(query).strip()
logging.info("Normalized query:\n%s", query)
mapper = TemplateMapper()
template = re.sub(r'({{[^{}]*}})', mapper, template)
logging.info("Template after replace:\n%s", template)
template = graphql.parse(template)
template = graphql.print_ast(template).strip()
logging.info("Template after normalized:\n%s", template)
template = '^' + re.escape(template) + '$'
for k, v in mapper.matches.items():
template = template.replace(k, v)
template = template.replace("\\ ", " ")
template = template.replace("\\\n", "\n")
task = RenderingTask(PYBARS, template, inject_methods=[reg_ex], inject_objects={"scope": None, 'key': None})
template, context = task.render_handlebars()
logging.info("Template after render:\n%s", template)
return re.match(template, query)
def test_fragments(ds):
query = """fragment NameAndAppearances on Character {
name
appearsIn
}
{
hero {
...NameAndAppearances
}
}"""
name_and_appearances = (
DSLFragment("NameAndAppearances")
.on(ds.Character)
.select(ds.Character.name, ds.Character.appearsIn)
)
query_dsl = DSLQuery(ds.Query.hero.select(name_and_appearances))
document = dsl_gql(name_and_appearances, query_dsl)
print(print_ast(document))
assert query == print_ast(document)
def test_get_introspection_query_ast(option):
introspection_query = get_introspection_query(
descriptions=option,
specified_by_url=option,
directive_is_repeatable=option,
schema_description=option,
)
dsl_introspection_query = get_introspection_query_ast(
descriptions=option,
specified_by_url=option,
directive_is_repeatable=option,
schema_description=option,
)
assert print_ast(gql(introspection_query)) == print_ast(dsl_introspection_query)
def print_query_plan(query_plan_descriptor: QueryPlanDescriptor,
indentation_depth: int = 4) -> str:
"""Return a string describing query plan."""
query_plan_strings = [""]
plan_and_depth = _get_plan_and_depth_in_dfs_order(
query_plan_descriptor.root_sub_query_plan)
for query_plan, depth in plan_and_depth:
line_separation = "\n" + " " * indentation_depth * depth
query_plan_strings.append(line_separation)
query_str = 'Execute in schema named "{}":\n'.format(
query_plan.schema_id)
query_str += print_ast(query_plan.query_ast)
query_str = query_str.replace("\n", line_separation)
query_plan_strings.append(query_str)
query_plan_strings.append("\n\nJoin together outputs as follows: ")
query_plan_strings.append(
str(query_plan_descriptor.output_join_descriptors))
query_plan_strings.append("\n\nRemove the following outputs at the end: ")
query_plan_strings.append(
str(query_plan_descriptor.intermediate_output_names) + "\n")
return "".join(query_plan_strings)
def test_dsl_root_type_not_default():
from graphql import parse, build_ast_schema
schema_str = """
schema {
query: QueryNotDefault
}
type QueryNotDefault {
version: String
}
"""
type_def_ast = parse(schema_str)
schema = build_ast_schema(type_def_ast)
ds = DSLSchema(schema)
query = dsl_gql(DSLQuery(ds.QueryNotDefault.version))
expected_query = """
{
version
}
"""
assert print_ast(query) == expected_query.strip()
with pytest.raises(GraphQLError) as excinfo:
DSLSubscription(ds.QueryNotDefault.version)
assert (
"Invalid field for <DSLSubscription>: <DSLField QueryNotDefault::version>"
) in str(excinfo.value)
async def _send_query(
self,
document: DocumentNode,
variable_values: Optional[Dict[str, str]] = None,
operation_name: Optional[str] = None,
) -> int:
"""Send a query to the provided websocket connection.
We use an incremented id to reference the query.
Returns the used id for this query.
"""
query_id = self.next_query_id
self.next_query_id += 1
payload: Dict[str, Any] = {"query": print_ast(document)}
if variable_values:
payload["variables"] = variable_values
if operation_name:
payload["operationName"] = operation_name
query_str = json.dumps({
"id": str(query_id),
"type": "start",
"payload": payload
})
await self._send(query_str)
return query_id
async def _send_query(
self,
document: DocumentNode,
variable_values: Optional[Dict[str, str]] = None,
operation_name: Optional[str] = None,
) -> int:
"""Send a query to the provided websocket connection.
We use an incremented id to reference the query.
Returns the used id for this query.
"""
query_id = self.next_query_id
self.next_query_id += 1
query_str = json.dumps({
"topic": self.channel_name,
"event": "doc",
"payload": {
"query": print_ast(document),
"variables": variable_values or {},
},
"ref": query_id,
})
await self._send(query_str)
return query_id
def test_use_variable_definition_multiple_times(ds):
var = DSLVariableDefinitions()
# `episode` variable is used in both fields
op = DSLMutation(
ds.Mutation.createReview.alias("badReview")
.args(review=var.badReview, episode=var.episode)
.select(ds.Review.stars, ds.Review.commentary),
ds.Mutation.createReview.alias("goodReview")
.args(review=var.goodReview, episode=var.episode)
.select(ds.Review.stars, ds.Review.commentary),
)
op.variable_definitions = var
query = dsl_gql(op)
assert (
print_ast(query)
== """mutation ($badReview: ReviewInput, $episode: Episode, $goodReview: ReviewInput) {
badReview: createReview(review: $badReview, episode: $episode) {
stars
commentary
}
goodReview: createReview(review: $goodReview, episode: $episode) {
stars
commentary
}
}"""
)
def test_multiple_operations(ds):
query = dsl_gql(
GetHeroName=DSLQuery(ds.Query.hero.select(ds.Character.name)),
CreateReviewMutation=DSLMutation(
ds.Mutation.createReview.args(episode=6,
review={
"stars":
5,
"commentary":
"This is a great movie!"
}).select(ds.Review.stars,
ds.Review.commentary)),
)
assert (print_ast(query) == """query GetHeroName {
hero {
name
}
}
mutation CreateReviewMutation {
createReview(episode: JEDI, review: {stars: 5, \
commentary: "This is a great movie!"}) {
stars
commentary
}
}
""")
async def execute(
self,
document: DocumentNode,
variable_values: Optional[Dict[str, str]] = None,
operation_name: Optional[str] = None,
extra_args: Dict[str, Any] = {},
) -> ExecutionResult:
"""Execute the provided document AST against the configured remote server.
This uses the aiohttp library to perform a HTTP POST request asynchronously
to the remote server.
The result is sent as an ExecutionResult object.
"""
query_str = print_ast(document)
payload: Dict[str, Any] = {
"query": query_str,
}
if variable_values:
payload["variables"] = variable_values
if operation_name:
payload["operationName"] = operation_name
post_args = {
"json": payload,
}
# Pass post_args to aiohttp post method
post_args.update(extra_args)
if self.session is None:
raise TransportClosed("Transport is not connected")
async with self.session.post(self.url, ssl=self.ssl,
**post_args) as resp:
try:
result = await resp.json()
except Exception:
# We raise a TransportServerError if the status code is 400 or higher
# We raise a TransportProtocolError in the other cases
try:
# Raise a ClientResponseError if response status is 400 or higher
resp.raise_for_status()
except ClientResponseError as e:
raise TransportServerError from e
raise TransportProtocolError(
"Server did not return a GraphQL result")
if "errors" not in result and "data" not in result:
raise TransportProtocolError(
"Server did not return a GraphQL result")
return ExecutionResult(errors=result.get("errors"),
data=result.get("data"))
def test_operation_name(ds):
query = dsl_gql(
GetHeroName=DSLQuery(ds.Query.hero.select(ds.Character.name), ))
assert (print_ast(query) == """query GetHeroName {
hero {
name
}
}
""")
def set_query_info(self, insert_var, query):
insert_var["query"] = {
"data": {
"name": query.name.value,
"query": graphql.print_ast(query)
},
"on_conflict": {
"constraint": "gql_query_query_key",
"update_columns": "query"
}
}
def build_func(provider, definition, schema, is_async):
"""
Builds a python function from a GraphQL AST definition
"""
name = definition.name.value
source = graphql.print_ast(definition)
assert definition.operation != graphql.OperationType.SUBSCRIPTION
params = [build_param(var) for var in definition.variable_definitions]
query_func = '__aquery__' if is_async else '__query__'
# TODO: Line numbers
if sys.version_info >= (3, 8):
py38 = {
'posonlyargs': [],
}
else:
py38 = {}
return ast.FunctionDef(
name=name,
args=ast.arguments(args=[],
defaults=[],
kwonlyargs=[
ast.arg(arg=name, annotation=None)
for name, _ in params
],
kw_defaults=[val for _, val in params],
vararg=None,
kwarg=None,
**py38),
body=[
ast.Return(value=ast.Call(
func=ast.Name(id=query_func, ctx=ast.Load()),
args=[
value2pyliteral(provider),
value2pyliteral(source),
],
keywords=[
ast.keyword(arg=name,
value=ast.Name(id=name, ctx=ast.Load()))
for name, _ in params
] + [
ast.keyword(arg=name,
value=(val if isinstance(val, ast.AST) else
value2pyliteral(val)))
for name, val in get_additional_kwargs(
provider, definition, schema).items()
],
), ),
],
decorator_list=[],
)
def test_subscription(ds):
query = dsl_gql(
DSLSubscription(
ds.Subscription.reviewAdded(episode=6).select(
ds.Review.stars, ds.Review.commentary)))
assert (print_ast(query) == """subscription {
reviewAdded(episode: JEDI) {
stars
commentary
}
}
""")
def tokenize_graphql_parameters(query):
"""
Parses a GraphQL query and replaces
parameter values with tokens to be used
for targeted injections
"""
document = parse(query)
for arg in document.definitions[0].selection_set.selections[0].arguments:
try:
if arg.value:
arg.value.value = "*"
except AttributeError:
error("Node did not have a parameter value to replace")
return print_ast(document)
def print_with_reduced_whitespace(ast: Document) -> str:
"""
Like the graphql-js print function, but deleting whitespace wherever
feasible. Specifically, all whitespace (outside of string literals) is
reduced to at most one space, and even that space is removed anywhere except
for between two alphanumerics.
"""
visit(ast, _HEX_CONVERSION_VISITOR)
val = re.sub(r'\s+', ' ', print_ast(ast))
val = re.sub(r'([^_a-zA-Z0-9]) ', _replace_with_first_group, val)
val = re.sub(r' ([^_a-zA-Z0-9])', _replace_with_first_group, val)
val = re.sub(r'"([a-f0-9]+)"', _from_hex, val)
return val
def input_value_info(graph, fields, ids):
nodes_map = _nodes_map(graph)
for ident in ids:
node = nodes_map[ident.node]
field = node.fields_map[ident.field]
option = field.options_map[ident.name]
if option.default is Nothing:
default = None
elif option.default is None:
# graphql-core currently can't parse/print "null" values
default = 'null'
else:
default = print_ast(ast_from_value(option.default))
info = {'id': ident,
'name': option.name,
'description': option.description,
'defaultValue': default}
yield [info[f.name] for f in fields]
def execution_node_for_group(
context: 'QueryPlanningContext',
group: 'FetchGroup',
parent_type: Optional[GraphQLCompositeType] = None,
) -> PlanNode:
selection_set = selection_set_from_field_set(group.fields, parent_type)
requires = (
selection_set_from_field_set(group.required_fields)
if len(group.required_fields) > 0
else None
)
variable_usages = context.get_variable_usages(selection_set, group.internal_fragments)
operation = (
operation_for_entities_fetch(selection_set, variable_usages, group.internal_fragments)
if requires is not None
else operation_for_root_fetch(
selection_set, variable_usages, group.internal_fragments, context.operation.operation
)
)
fetch_node = FetchNode(
service_name=group.service_name,
requires=trim_selection_nodes(requires.selections) if requires is not None else None,
variable_usages=list(variable_usages.keys()),
operation=strip_ignored_characters(print_ast(operation)),
)
node: PlanNode = (
FlattenNode(path=group.merge_at, node=fetch_node)
if group.merge_at is not None and len(group.merge_at) > 0
else fetch_node
)
if len(group.dependent_groups) > 0:
dependent_nodes = [
execution_node_for_group(context, dependent_group)
for dependent_group in group.dependent_groups
]
return flat_wrap('Sequence', [node, flat_wrap('Parallel', dependent_nodes)])
else:
return node
def test_add_variable_definitions_with_default_value_enum(ds):
var = DSLVariableDefinitions()
op = DSLMutation(
ds.Mutation.createReview.args(
review=var.review, episode=var.episode.default(4)
).select(ds.Review.stars, ds.Review.commentary)
)
op.variable_definitions = var
query = dsl_gql(op)
assert (
print_ast(query)
== """mutation ($review: ReviewInput, $episode: Episode = NEWHOPE) {
createReview(review: $review, episode: $episode) {
stars
commentary
}
}"""
)
async def _send_query(
self,
document: DocumentNode,
variable_values: Optional[Dict[str, Any]] = None,
operation_name: Optional[str] = None,
) -> int:
query_id = self.next_query_id
self.next_query_id += 1
data: Dict = {"query": print_ast(document)}
if variable_values:
data["variables"] = variable_values
if operation_name:
data["operationName"] = operation_name
serialized_data = json.dumps(data, separators=(",", ":"))
payload = {"data": serialized_data}
message: Dict = {
"id": str(query_id),
"type": "start",
"payload": payload,
}
assert self.auth is not None
message["payload"]["extensions"] = {
"authorization": self.auth.get_headers(serialized_data)
}
await self._send(json.dumps(
message,
separators=(",", ":"),
))
return query_id
def test_add_variable_definitions_with_default_value_input_object(ds):
var = DSLVariableDefinitions()
op = DSLMutation(
ds.Mutation.createReview.args(
review=var.review.default({"stars": 5, "commentary": "Wow!"}),
episode=var.episode,
).select(ds.Review.stars, ds.Review.commentary)
)
op.variable_definitions = var
query = dsl_gql(op)
assert (
print_ast(query)
== """
mutation ($review: ReviewInput = {stars: 5, commentary: "Wow!"}, $episode: Episode) {
createReview(review: $review, episode: $episode) {
stars
commentary
}
}""".strip()
)
def test_type_hero_query(ds):
query = """{
__type(name: "Hero") {
kind
name
ofType {
kind
name
}
}
}"""
type_hero = DSLMetaField("__type")(name="Hero")
type_hero.select(
ds.__Type.kind,
ds.__Type.name,
ds.__Type.ofType.select(ds.__Type.kind, ds.__Type.name),
)
query_dsl = DSLQuery(type_hero)
assert query == str(print_ast(dsl_gql(query_dsl))).strip()
def test_add_variable_definitions_in_input_object(ds):
var = DSLVariableDefinitions()
op = DSLMutation(
ds.Mutation.createReview.args(
review={"stars": var.stars, "commentary": var.commentary},
episode=var.episode,
).select(ds.Review.stars, ds.Review.commentary)
)
op.variable_definitions = var
query = dsl_gql(op)
assert (
print_ast(query)
== """mutation ($stars: Int, $commentary: String, $episode: Episode) {
createReview(
review: {stars: $stars, commentary: $commentary}
episode: $episode
) {
stars
commentary
}
}"""
)
def max_rps_test(self, query):
query_str = graphql.print_ast(query)
print(
Fore.GREEN +
"(Compute maximum Request per second) Running wrk benchmark for query\n",
query_str + Style.RESET_ALL)
self.hge.graphql_q(query_str) # Test query once for errors
bench_script = os.path.join(self.lua_dir + '/bench-wrk.lua')
graphql_url = self.hge.url + '/v1/graphql'
params = self.get_wrk2_params()
duration = 30
wrk_command = [
'wrk', '-t',
str(params['threads']), '-c',
str(params['connections']), '-d',
str(duration), '--latency', '-s', bench_script, graphql_url,
query_str
]
self.docker_client = docker.from_env()
result = self.docker_client.containers.run(
self.wrk_docker_image,
detach=False,
stdout=False,
stderr=True,
command=wrk_command,
network_mode='host',
environment=self.get_lua_env(),
volumes=self.get_scripts_vol(),
remove=True,
user=self.get_current_user())
summary = json.loads(result)['summary']
# TODO explain this calculation. Why aren't we using wrk's reported 'max'? Should we call this avg_sustained_rps or something?
max_rps = round(summary['requests'] / float(duration))
self.insert_max_rps_result(query, max_rps)
print("Max RPS", max_rps)
return max_rps
def print_query_plan(query_plan_descriptor, indentation_depth=4):
"""Return a string describing query plan."""
query_plan_strings = [u'']
plan_and_depth = _get_plan_and_depth_in_dfs_order(
query_plan_descriptor.root_sub_query_plan)
for query_plan, depth in plan_and_depth:
line_separation = u'\n' + u' ' * indentation_depth * depth
query_plan_strings.append(line_separation)
query_str = u'Execute in schema named "{}":\n'.format(
query_plan.schema_id)
query_str += print_ast(query_plan.query_ast)
query_str = query_str.replace(u'\n', line_separation)
query_plan_strings.append(query_str)
query_plan_strings.append(u'\n\nJoin together outputs as follows: ')
query_plan_strings.append(
str(query_plan_descriptor.output_join_descriptors))
query_plan_strings.append(u'\n\nRemove the following outputs at the end: ')
query_plan_strings.append(
str(query_plan_descriptor.intermediate_output_names) + u'\n')
return ''.join(query_plan_strings)
def _check_query_node_structure_helper(self, query_node,
expected_query_node):
"""Check query_node has the same structure as expected_query_node."""
# Check AST and id of the parent
self.assertEqual(print_ast(query_node.query_ast),
expected_query_node.query_str)
self.assertEqual(query_node.schema_id, expected_query_node.schema_id)
# Check number of children matches
child_query_connections = query_node.child_query_connections
expected_child_data = expected_query_node.child_query_nodes_and_out_names
self.assertEqual(len(child_query_connections),
len(expected_child_data))
for i, (child_query_connection,
expected_child_data_piece) in enumerate(
six.moves.zip(child_query_connections,
expected_child_data)):
# Check child and parent connections
child_query_node = child_query_connection.sink_query_node
child_expected_query_node, parent_out_name, child_out_name = expected_child_data_piece
self._check_query_node_edge(query_node, i, child_query_node,
parent_out_name, child_out_name)
# Recurse
self._check_query_node_structure_helper(child_query_node,
child_expected_query_node)
def root_resolver(obj, info, **inputs):
query = graphql.print_ast(info.operation)
# TODO: add graphql context
delegated_result = client.execute(query=query,
variables=info.variable_values)
return delegated_result.data[info.field_name]
def __str__(self):
return print_ast(self.ast_field)
async def execute(
self,
document: DocumentNode,
variable_values: Optional[Dict[str, str]] = None,
operation_name: Optional[str] = None,
extra_args: Dict[str, Any] = None,
upload_files: bool = False,
) -> ExecutionResult:
"""Execute the provided document AST against the configured remote server
using the current session.
This uses the aiohttp library to perform a HTTP POST request asynchronously
to the remote server.
Don't call this coroutine directly on the transport, instead use
:code:`execute` on a client or a session.
:param document: the parsed GraphQL request
:param variables_values: An optional Dict of variable values
:param operation_name: An optional Operation name for the request
:param extra_args: additional arguments to send to the aiohttp post method
:param upload_files: Set to True if you want to put files in the variable values
:returns: an ExecutionResult object.
"""
query_str = print_ast(document)
payload: Dict[str, Any] = {
"query": query_str,
}
if operation_name:
payload["operationName"] = operation_name
if upload_files:
# If the upload_files flag is set, then we need variable_values
assert variable_values is not None
# If we upload files, we will extract the files present in the
# variable_values dict and replace them by null values
nulled_variable_values, files = extract_files(
variables=variable_values, file_classes=self.file_classes,
)
# Save the nulled variable values in the payload
payload["variables"] = nulled_variable_values
# Prepare aiohttp to send multipart-encoded data
data = aiohttp.FormData()
# Generate the file map
# path is nested in a list because the spec allows multiple pointers
# to the same file. But we don't support that.
# Will generate something like {"0": ["variables.file"]}
file_map = {str(i): [path] for i, path in enumerate(files)}
# Enumerate the file streams
# Will generate something like {'0': <_io.BufferedReader ...>}
file_streams = {str(i): files[path] for i, path in enumerate(files)}
# Add the payload to the operations field
operations_str = json.dumps(payload)
log.debug("operations %s", operations_str)
data.add_field(
"operations", operations_str, content_type="application/json"
)
# Add the file map field
file_map_str = json.dumps(file_map)
log.debug("file_map %s", file_map_str)
data.add_field("map", file_map_str, content_type="application/json")
# Add the extracted files as remaining fields
for k, v in file_streams.items():
data.add_field(k, v, filename=k)
post_args: Dict[str, Any] = {"data": data}
else:
if variable_values:
payload["variables"] = variable_values
if log.isEnabledFor(logging.INFO):
log.info(">>> %s", json.dumps(payload))
post_args = {"json": payload}
# Pass post_args to aiohttp post method
if extra_args:
post_args.update(extra_args)
if self.session is None:
raise TransportClosed("Transport is not connected")
async with self.session.post(self.url, ssl=self.ssl, **post_args) as resp:
try:
result = await resp.json()
if log.isEnabledFor(logging.INFO):
result_text = await resp.text()
log.info("<<< %s", result_text)
except Exception:
# We raise a TransportServerError if the status code is 400 or higher
# We raise a TransportProtocolError in the other cases
try:
# Raise a ClientResponseError if response status is 400 or higher
resp.raise_for_status()
except ClientResponseError as e:
raise TransportServerError(str(e)) from e
result_text = await resp.text()
raise TransportProtocolError(
f"Server did not return a GraphQL result: {result_text}"
)
if "errors" not in result and "data" not in result:
result_text = await resp.text()
raise TransportProtocolError(
"Server did not return a GraphQL result: "
'No "data" or "error" keys in answer: '
f"{result_text}"
)
return ExecutionResult(errors=result.get("errors"), data=result.get("data"))
def wrk2_test(self, query, rps):
def upload_files(files):
if self.upload_root_uri:
p = urlparse(self.upload_root_uri)
if p.scheme == 's3':
bucket = p.netloc
key = p.path.lstrip('/')
s3_client = boto3.client('s3')
for (f, f_key) in files:
s3_client.upload_file(f, bucket,
os.path.join(key, f_key))
query_str = graphql.print_ast(query)
params = self.get_wrk2_params()
print(
Fore.GREEN +
"Running benchmark wrk2 for at {} req/s (duration: {}) for query\n"
.format(rps, params['duration']), query_str + Style.RESET_ALL)
bench_script = os.path.join(self.lua_dir, 'bench-wrk2.lua')
graphql_url = self.hge.url + '/v1/graphql'
timestamp = datetime.datetime.now().strftime('%Y-%m-%d_%H-%M-%S')
results_dir = self.results_root_dir
tests_path = [str(rps), timestamp]
results_dir = os.path.join(results_dir, *tests_path)
os.makedirs(results_dir, exist_ok=True)
wrk2_command = [
'wrk2', '-R',
str(rps), '-t',
str(params['threads']), '-c',
str(params['connections']), '-d',
str(params['duration']), '--latency', '-s', bench_script,
graphql_url, query_str, results_dir
]
volumes = self.get_scripts_vol()
volumes[results_dir] = {'bind': results_dir, 'mode': 'rw'}
self.docker_client = docker.from_env()
result = self.docker_client.containers.run(
self.wrk_docker_image,
detach=False,
stdout=True,
stderr=False,
command=wrk2_command,
network_mode='host',
environment=self.get_lua_env(),
volumes=volumes,
remove=True,
user=self.get_current_user()).decode('ascii')
histogram_file = os.path.join(results_dir, 'latencies.hgrm')
histogram = self.get_latency_histogram(result, histogram_file)
summary_file = os.path.join(results_dir, 'summary.json')
with open(summary_file) as f:
summary = json.load(f)
latencies_file = os.path.join(results_dir, 'latencies')
def extract_data(v):
return v['data'] if isinstance(v, dict) and 'data' in v else v
tests_info = {
k: extract_data(v)
for (k, v) in self.gen_test_info(query, rps).items()
}
tests_setup_file = os.path.join(results_dir, 'test_setup.json')
with open(tests_setup_file, 'w') as f:
f.write(json.dumps(tests_info, indent=2))
upload_files([(x, os.path.join(*tests_path, y)) for (x, y) in [(
summary_file, 'summary.json'), (
latencies_file, 'latencies'), (
histogram_file,
'latencies.hgrm'), (tests_setup_file, 'test_setup.json')]])
if self.upload_root_uri:
latencies_uri = uri_path_join(self.upload_root_uri, *tests_path,
'latencies')
else:
latencies_uri = pathlib.Path(latencies_file).as_uri()
self.insert_result(query, rps, summary, histogram, latencies_uri)
return (summary, histogram)