When "Set View" is clicked on, there is no action on large datasets. The delay should be able to be avoided.

the current implementation of annotation management is based from the desktop version and is fundamentally improper (if barely correct)

1.Open "Display Preference" and click on "Annotations", the view does not change by setting.
2.Click on "gene symbol/marker id", the view also does not change by setting.
3. click on "Reset", the tabular view is not reset to original.

When change code on tabularview component. Please be careful that the change may break menu selector functions. After change it please not only test on tabular view, but also test on menu selector functions.

There are two constructors of AnalysisSubmissionEvent: one takes dataset ID only; the other takes bison DSDataSet. The latter is supposed to be replaced, but at this time used more - 5 places. This is related to issue #10.

org.geworkbenchweb.plugins.markus.MarkusAnalysis.execute(75) reports FINISHED when it is in fact pending.

Currently, the dataset, mainly just the microarray dataset, is first read and parsed to create a bison object and serialized; when the data is needed, the bison object is deserialized. That does not add value compared with parsing the original file again, but add a major layer of complexity and inefficiency.

The implementation consists of serializeDataSet and deserializedDataSet methods in UserDirUtils. serializeDataSet is referenced in one place and deserialzeDataSet is referenced in many places (10). So it is better to replace the deserializeDataSet one by one first.

The new data flow is basically to replace the serialized object with two mechanisms: part of the data is serialized in JPA to make it more efficient to query and cleaner to manage; at the same time, the original data file is explicitly reserved so we can parse it again when we need to access data not included in the current persistence data schema.

This issue is the core code, but directly affects the annotation revamp #3 in this project.
InteractionsConnectionImpl.getInteractionsByEntrezIdOrGeneSymbol_1 and getInteractionsByEntrezIdOrGeneSymbol_2 depend on the behaviors of DSGeneMarker.getGeneName() and DSGeneMarker.getGeneName(), which are decided by the 'current dataset' because that is how annotation is managed. 'Current dataset' is open to change outside of annotation mechanism and CNKB.

Instead, this could be made reliable just by using gene symbol and gene ID directly, not the DSGeneMarker objects.

It has never turned out that the catch-all exception clause is the proper solution for the problem. It is a very bad idea. When it swallows the exception and does output or log anything, it is worse. In a case in AnalysisListener, it does not swallow the exception, but allows the execution to continue - that is not the right thing to do either.

vaadin 6 is out of date. It is getting harder to get supports in documentation and tools.

CNKB - create network incorrectly when interactom "HGi" is selected

The recent implementation of node renaming feature caused a strange problem of vaddin: the LAST data note's children result nodes do not show up; consequently, a lot of odd behavior occurs after the result nodes are supposed to be created for the last data node. Without the node to add action handler (for renaming), everything works fine.

When you click on "tabluar view" from the menu page for microarray dataset, it takes a few seconds to bring up the first page (25 rows) of the tabular view.

open filter window from menu, when array sets are selected and submitted, the data displayed on tabular view is wrong.

The combo box to choose file type is used in the desktop version to filter out the files by the desired type. In the web version, the filtering is not implemented, so this combo box serves no purpose and add an unnecessary step to operate. It should be removed until we implement a filtering mechanism.

When selected gene is not in annotation list, CNKB will get server response code 500

In search box, type a character, exception is thrown: (I switched to a much earlier commit 3242017 and it has the same issue, so it is not caused by the recent data-flowing change)

Caused by: java.lang.IllegalArgumentException: PagedTable can only use containers that implement Container.Indexed
at org.geworkbenchweb.utils.PagedTable.setContainerDataSource(PagedTable.java:236)
at org.geworkbenchweb.plugins.TableMenuSelector$3$1.textChange(TableMenuSelector.java:103)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:606)
at com.vaadin.event.ListenerMethod.receiveEvent(ListenerMethod.java:510)
... 28 more

So the building and deployment process will be:

• more consistent
• better documented, especially for the library dependency
• more efficient

Although the requested feature overlaps some of the functionality of genspace component, the purpose is different from that of genspace, and it is better to design this as a separate element in the system.

The activities to be logged include: logging-in, logging-out, loading a dataset, starting an analysis, and a result set having been created.

CNKBUI.java needs to get HttpSession directly to support its functionality of "CNKB user" authorization. The implementation had problem of getting null pointer in the past. The problem disappeared later and the cause has not been found out. I saw the same issue again (CNKBUI.java line 303 in method getInteractions) when I am working on other changes.

After I register a new user, it shows a warning (cannot re-create user). After dismissing the message, it works fine.

The annotation files under public/private should not use file name as item id. It should read from database and use annotation file id as item id. Also the application should not base on file disk to determine if a private or public annotation file exists or not, especially we deploy geworkbench-web to production which is a distributed server. An individual program will be developed to insert a public annotation file to database.

The current implementation chose to freeze the application during potentially long data loading. This is problematic. Instead, it should disable certain part of functionality - even more parts than strictly necessary is OK - and leave the application responsive.

UploadDataUI.java

the value of gene type and annotation are not populated when a query is performed.

This is a requirement of CPTAC enhancement.

The main change is to add interaction type to the network data model.

For the specification of the SIF format, see http://wiki.c2b2.columbia.edu/workbench/index.php/Cellular_Networks_KnowledgeBase#SIF_format

This is similar to, or arguably part of, issue #10, but is used more at this time and does not necessarily have anything to do with the bison data model.

According to the admin group's report, there are 12 vulnerable files in this application.

Tabular view - Preference "precision" setting does not work properly. When you change precision setting, only one page return back to tabular view.

These two components depend on invient chart,which is no longer available for free in vaadin 7.
Two promising candidates to replace invient chart are:

1. https://vaadin.com/directory#addon/jfreechart-wrapper-for-vaadin
2. https://vaadin.com/directory#addon/freecode-charts-and-d3-wrapper

jfreechat will be perfect if the interactive behavior (tooltip etc) can be worked out easily because this is the same library we are using in the swing version.

The current parsing code (from desktop application) is tightly coupled with bison data types and causes unnecessary inefficiency. It also includes some undocumented and likely unintended logic. The new code will handle parsing in a self contained way, and provides conversion mechanism to bison data type for backward compatibility.

It was used before the current set loading exits in the set view. It causes unnecessary complexity and is not consistent with other analysis tools.

Currently, the latest obo file is downloaded when the application is loaded the first time. The download happens in a background thread started from the static block of class GeneOntologyTree.

• Pro: the latest file at the time of GeneOntologyTree being class-loaded is used.
• Con: 1) The multi-thread model of coordinating the downloading thread and the code using GennOntologyTree, including the fallback when the download fails, is complicated. 2) The server providing the obo file occasionally (though not very often) does not respond normally - instead of falling back to the existing copy quickly, the application may end up to have a corrupted copy of obo file that appears ready to be used.

Summarily, the benefit of the current model is minimal, and it is not worth of the consequent complexity that has not been totally taken care of. I think we should use a simple process that is separated from the web application to update the obo file when necessary (manually or automatically).

The throttle graph of CNKB result is not properly redrawn upon split pane being moved by the bottom table part. The Threshold slider also moves the split pane and I think it should not.

Gene Ontology Tree are populated in afdev server, but not in production server.

You can see detail on web1 /var/log/tomcat/geworkbench.c2b2.columbia.edu.log

Marker selector shows all the markers. In case of bcell100.exp, it shows all 12600 markers for the user to choose. This would be better to show the marker list in smaller pages.