Building the table for speed releasing.

dashboard/dashboard/models/table_config.py

Index: dashboard/dashboard/models/table_config.py
diff --git a/dashboard/dashboard/models/table_config.py b/dashboard/dashboard/models/table_config.py
index d609bc1f1d946b30759b3ff36dd6f681a7e8f3e9..edd6e6e247235ec59b37797303605544d576b451 100644
--- a/dashboard/dashboard/models/table_config.py
+++ b/dashboard/dashboard/models/table_config.py
@@ -31,6 +31,7 @@ import json
 
 from google.appengine.ext import ndb
 
+from dashboard.common import utils
 from dashboard.models import internal_only_model
 
 
@@ -96,6 +97,13 @@ def CreateTableConfig(name, bots, tests, layout, username):
     error_message = '%s already exists.' % name
     return None, error_message
 
+  for bot in bots:
+    for test in tests:
+      test_key = utils.TestMetadataKey(bot + '/' + test)
+      if not test_key.get():
+        error_message = '%s is not a valid test.' % test
+        return None, error_message

[eakuefner, 2017/01/28 00:22:17]

It occurs to me that this pattern is kind of funky. Can you throw exceptions in
these cases instead of having this function return a tuple?

This is an example of a place where you would throw a BadRequestError, and then
let it propagate up to the handler, which can catch it and do self.ReportError.
Check out how add_point does this:
https://github.com/catapult-project/catapult/blob/master/dashboard/dashboard/...

[jessimb, 2017/01/30 19:23:10]

Feels much cleaner!
Note that I didn't use self.ReportError because simple_xhr will display the Http
status code if it's not 200 and I want it to display the message regardless of
the status. Instead, I still added the logging in and
self.response.out.write(the error message). This way the error is recorded on
the server and the user knows how to fix their inputs so it can actually
validate the next time.

+
   # Input validates, create table now.
   error_message = 'TableConfig created successfully.'
   table_config = TableConfig(id=name, bots=valid_bots, tests=tests,