Project

General

Profile

Bug #1059

create testing harness for ecogrid

Added by Matt Jones over 17 years ago. Updated almost 15 years ago.

Status:
New
Priority:
Immediate
Assignee:
Category:
ecogrid
Target version:
Start date:
05/19/2003
Due date:
% Done:

0%

Estimated time:
Bugzilla-Id:
1059

Description

We need a common testing harness that can be used to launch client queries
against ecogrid nodes and validate that the results returned are correct. This
system would be a basic client application, and so the core of it could be
reused in other clients if it is properly designed.

Who: Higgins
Target: June 9

History

#1 Updated by Matt Jones almost 17 years ago

Some more detail has emerged on this from the Santa Barbara meeting. The goal
of this bug is to create a series of unit tests that test the published EcoGrid
APIs for each of the wrapped services (e.g., metacat/srb/DiGIR) to make sure
that they all implement the interfaces uniformly. This can be implemented in
phases, as the EcoGrip APIs are implemented. So for example, because query APIs
are being developed before data insertion APIs, it is ok to assume for now that
data being queried is inserted through a non-ecogrid mechanism. Once the write
APIs are established, the test should login, insert the test data, query the
test data, modify and delete the test data, and test exceptional conditions
(such as trying to insert after an invalid login, trying to delete/move/change
data owned by another user, etc).

Running these tests should be accomplished through the ant build system by
typing "ant test" on the commandline. We have some existing ant build files
that run tests, so see me about how to set this up if you have questions. It
would be best if the build file had a configurable parameter that lists the
implementations to be tested (ie, list a metacat, srb, digir, etc implementation
and run through each of those with all of the tests).

Some minimum steps needed to close this bug. It is not a complete list.
Individual parts may be broken out into their own individual bugs if that makes
it easier to track partial completion of the task.

1) Get test data and queries
a) Obtain representative EML and Darwin Core metadata and data, place in CVS
at seek/projects/ecogrid/test/testfiles in a suitably documented way
b) Rewrite and extend the test queries already started in CVS to be a full
suite of query test documents that test all of the features of the query language
2) Write a JUnit test suite that is configurable to point at specific
implementations of the EcoGridAPIs. This might be
tests/org.ecoinformatics.seek.ecogridtest.EcoGridAPITest.java
3) Have a separate function for testing each of the following functions:
a) insert metadata
b) insert data
c) modify metadata
d) modify data
e) query metadata -- get the right set of results (e.g., query method)
f) download data -- using an identifer (e.g., "get" method)
g) query data -- download a subset of data
h) delete data
i) register an ecogrid node
j) query registry for nodes
k) login (authenticate)
l) logout
m) test access control restrictions

Each of these involves several possible EcoGrid API method invocations.
Sometimes more than one way to do it will exist (e.g., EcoGrid Query Level I and
Level II interfaces) -- we need to test all legitimate approaches, and make sure
illegitimate approaches are blocked.

As I said, these will probably not be developed in this order. The first
methods to test will probably be the EcoGrid Level I query API, basically the
search and get methods, as described in bug 1041.

Developing these tests will have some beneficial side-impacts. First, it will
clarify our thinking about the APIs and make sure that they allow the
expressiveness we want. Second, it will provide a client library (if properly
designed) that can be used in other contexts. Third, it will provide immediate
feedback to developers doing wrapping of new services as to whether they have
gotten it right or not.

#2 Updated by Matt Jones almost 15 years ago

This still needs to be done before the release can occur. An additional design
goal is to stress test with mutiple simultaneous client connections (probably a
configurable number of client connections) to make sure the implementations can
handle getting hit with many requests at the same time.

#3 Updated by Redmine Admin over 7 years ago

Original Bugzilla ID was 1059

Also available in: Atom PDF