Fix serve dependencies (#9192)

python/ray/serve/examples/echo.py

@@ -2,12 +2,24 @@
 Example service that prints out http context.
 """
 
+import json
 import time
 
+from pygments import formatters, highlight, lexers
+
 import requests
 
 from ray import serve
-from ray.serve.utils import pformat_color_json
+
+
+def pformat_color_json(d):
+    """Use pygments to pretty format and colorize dictionary"""
+    formatted_json = json.dumps(d, sort_keys=True, indent=4)
+
+    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
+                              formatters.TerminalFormatter())
+
+    return colorful_json
 
 
 def echo(flask_request):

python/ray/serve/examples/echo_actor.py

@@ -5,13 +5,25 @@ come from either web (parsing Flask request) or python call.
 This actor can be called from HTTP as well as from Python.
 """
 
+import json
 import time
 
+from pygments import formatters, highlight, lexers
+
 import requests
 
 import ray
 from ray import serve
-from ray.serve.utils import pformat_color_json
+
+
+def pformat_color_json(d):
+    """Use pygments to pretty format and colorize dictionary"""
+    formatted_json = json.dumps(d, sort_keys=True, indent=4)
+
+    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
+                              formatters.TerminalFormatter())
+
+    return colorful_json
 
 
 class MagicCounter:

python/ray/serve/examples/echo_actor_batch.py

@@ -5,13 +5,25 @@ The queries incoming to this actor are batched.
 This actor can be called from HTTP as well as from Python.
 """
 
+import json
 import time
 
+from pygments import formatters, highlight, lexers
+
 import requests
 
 import ray
 from ray import serve
-from ray.serve.utils import pformat_color_json
+
+
+def pformat_color_json(d):
+    """Use pygments to pretty format and colorize dictionary"""
+    formatted_json = json.dumps(d, sort_keys=True, indent=4)
+
+    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
+                              formatters.TerminalFormatter())
+
+    return colorful_json
 
 
 class MagicCounter:

python/ray/serve/examples/echo_error.py

@@ -13,13 +13,25 @@ This shows that error is hidden from HTTP side but always visible when calling
 from Python.
 """
 
+import json
 import time
 
+from pygments import formatters, highlight, lexers
+
 import requests
 
 import ray
 from ray import serve
-from ray.serve.utils import pformat_color_json
+
+
+def pformat_color_json(d):
+    """Use pygments to pretty format and colorize dictionary"""
+    formatted_json = json.dumps(d, sort_keys=True, indent=4)
+
+    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
+                              formatters.TerminalFormatter())
+
+    return colorful_json
 
 
 def echo(_):

python/ray/serve/examples/echo_fixed_packing.py

@@ -1,47 +0,0 @@
-"""
-Example showing fixed packing policy. The outputs from
-v1 and v2 will be coming according to packing_num specified!
-This is a packed round robin example. First batch of packing_num
-(five in this example) queries would go to 'echo:v1' backend and
-then next batch of packing_num queries would go to 'echo:v2'
-backend.
-"""
-import time
-
-import requests
-
-from ray import serve
-from ray.serve.utils import pformat_color_json
-
-
-def echo_v1(_):
-    return "v1"
-
-
-def echo_v2(_):
-    return "v2"
-
-
-# specify the router policy as FixedPacking with packing num as 5
-serve.init(
-    queueing_policy=serve.RoutePolicy.FixedPacking,
-    policy_kwargs={"packing_num": 5})
-
-# create first backend
-serve.create_backend("echo:v1", echo_v1)
-
-# create service backed by the first backend
-serve.create_endpoint("my_endpoint", backend="echo:v1", route="/echo")
-
-# create second backend
-serve.create_backend("echo:v2", echo_v2)
-
-# split the service between the two backends
-serve.set_traffic("my_endpoint", {"echo:v1": 0.5, "echo:v2": 0.5})
-
-while True:
-    resp = requests.get("http://127.0.0.1:8000/echo").json()
-    print(pformat_color_json(resp))
-
-    print("...Sleeping for 2 seconds...")
-    time.sleep(2)

python/ray/serve/examples/echo_full.py

@@ -2,13 +2,26 @@
 Full example of ray.serve module
 """
 
+import json
 import time
 
+from pygments import formatters, highlight, lexers
+
 import requests
 
 import ray
 import ray.serve as serve
-from ray.serve.utils import pformat_color_json
+
+
+def pformat_color_json(d):
+    """Use pygments to pretty format and colorize dictionary"""
+    formatted_json = json.dumps(d, sort_keys=True, indent=4)
+
+    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
+                              formatters.TerminalFormatter())
+
+    return colorful_json
+
 
 # initialize ray serve system.
 serve.init()

python/ray/serve/examples/echo_round_robin.py

@@ -1,41 +0,0 @@
-"""
-Example showing round robin policy. The outputs from
-v1 and v2 will be (almost) interleaved as queries get processed.
-"""
-import time
-
-import requests
-
-from ray import serve
-from ray.serve.utils import pformat_color_json
-
-
-def echo_v1(_):
-    return "v1"
-
-
-def echo_v2(_):
-    return "v2"
-
-
-# specify the router policy as RoundRobin
-serve.init(queueing_policy=serve.RoutePolicy.RoundRobin)
-
-# create first backend
-serve.create_backend("echo:v1", echo_v1)
-
-# create a service backend by the first backend
-serve.create_endpoint("my_endpoint", backend="echo:v1", route="/echo")
-
-# create second backend
-serve.create_backend("echo:v2", echo_v2)
-
-# split the service between the two backends
-serve.set_traffic("my_endpoint", {"echo:v1": 0.5, "echo:v2": 0.5})
-
-while True:
-    resp = requests.get("http://127.0.0.1:8000/echo").json()
-    print(pformat_color_json(resp))
-
-    print("...Sleeping for 2 seconds...")
-    time.sleep(2)

python/ray/serve/examples/echo_split.py

@@ -2,12 +2,24 @@
 Example of traffic splitting. We will first use echo:v1. Then v1 and v2
 will split the incoming traffic evenly.
 """
+import json
 import time
 
+from pygments import formatters, highlight, lexers
+
 import requests
 
 from ray import serve
-from ray.serve.utils import pformat_color_json
+
+
+def pformat_color_json(d):
+    """Use pygments to pretty format and colorize dictionary"""
+    formatted_json = json.dumps(d, sort_keys=True, indent=4)
+
+    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
+                              formatters.TerminalFormatter())
+
+    return colorful_json
 
 
 def echo_v1(_):

python/ray/serve/utils.py

@@ -10,7 +10,6 @@ import io
 import os
 
 import requests
-from pygments import formatters, highlight, lexers
 
 import ray
 from ray.serve.constants import HTTP_PROXY_TIMEOUT
@@ -79,16 +78,6 @@ class ServeEncoder(json.JSONEncoder):
         return super().default(o)
 
 
-def pformat_color_json(d):
-    """Use pygments to pretty format and colroize dictionary"""
-    formatted_json = json.dumps(d, sort_keys=True, indent=4)
-
-    colorful_json = highlight(formatted_json, lexers.JsonLexer(),
-                              formatters.TerminalFormatter())
-
-    return colorful_json
-
-
 def block_until_http_ready(http_endpoint,
                            backoff_time_s=1,
                            timeout=HTTP_PROXY_TIMEOUT):

python/setup.py

@@ -83,7 +83,7 @@ if "RAY_USE_NEW_GCS" in os.environ and os.environ["RAY_USE_NEW_GCS"] == "on":
 extras = {
     "debug": [],
     "dashboard": ["requests", "gpustat"],
-    "serve": ["uvicorn", "flask", "blist"],
+    "serve": ["uvicorn", "flask", "blist", "requests"],
     "tune": ["tabulate", "tensorboardX", "pandas"]
 }