g-node / g-node-portal Goto Github PK

how the ID of the object should look like:

• integer
• globally unique, like hash
• etc.

how to represent a link to an object:

• including hostname, like http://portal.g-node.org/data/neo/<obj_type>/<obj_id>
• just the relative URL say /neo/<obj_type>/<obj_id>
• host + hash, like http://portal.g-node.org/data/<obj_hash>/
• etc.

• if a record was uploaded and one check failed, i updated it with the
  correct files. so far the progress bar did not disappear. something
  got stuck?
• add record and update record dialogs can be open at the same time

Should we show in the REST response reverse connections via related name? Means, not like 'property_set' but like 'properties'?

Uploading an evaluation file for Espens Benchmark data creates a new evaluation that does not come out of the "in progress" state.

How to reprocude:
Upload the attached file for Espens Benchmark (unfiltered).
How do I attach a file? The file is now in the dropbox: \Dropbox\OsloDataShare\espen_websitetest_sorted.gdf

(maybe a little help icon with a description text?)

This defines the immediate goal of the development. I did not find how to put that here, so it is an issue.

Goal: Bring the website to a state where it is fully operational and can be presented in a way that the evaluation results to one benchmark are understandable and presentable.
Important:

1. The benchmark upload must work for files that are according to the specification.
2. The evaluation must work for files that are according to the specification.
3. The evaluation results must be accessible.
4. The evaluation results must be presented in a "human" readable and at least rudimentary nice way.
5. The "evaluation view filters", i.e. the summary presentation e.g. of all evaluations of one benchmark must work.
6. One useful example benchmark must be uploaded.
7. Several useful sortings must be uploaded for that example benchmark.

The scripts for creating figures need to be done in a nicer manner.

explanation needed or an icon?

When uploading a sorting for an evaluation the user should specify what kind of method he used. This is so far done in the field "algorithm". Since it would be very interesting for another user to filter all evaluations according to the type of method that was used, we need to specify that as well. Type of method I mean:

using ground truth templates, manual, semi automated, fully automated

Subject says it all.

When creating a benchmark record, you can specify if you want the groundtruth to be downloadable.

1. change the text label from "GT" to groundtrurh
2. move the field 2 rows down so that is is next to the "select groundtruth" file

Traceback (most recent call last):

File "/opt/g-node-portal/apps/lib/python2.5/site-packages/django/core/handlers/base.py", line 111, in get_response
response = callback(request, _callback_args, *_callback_kwargs)

File "/opt/g-node-portal/apps/lib/python2.5/site-packages/django/contrib/auth/decorators.py", line 23, in _wrapped_view
return view_func(request, _args, *_kwargs)

File "/opt/g-node-portal/apps/g-node-portal/apps/account/views.py", line 179, in password_change
password_change_form.save()

File "/opt/g-node-portal/apps/g-node-portal/apps/account/forms.py", line 303, in save
if not (change == True):

UnboundLocalError: local variable 'change' referenced before assignment

<WSGIRequest
GET:<QueryDict: {}>,
POST:<QueryDict: {u'action': [u'change password'], u'password1': [u'xxHIDDENxx'], u'password2': [u'xxHIDDENxx'], u'oldpassword': [u'xxHIDDENxx']}>,
COOKIES:{'sessionid': 'd63df70e77b7b6dcbe4287c82f72ba1c'},
META:{'CONTENT_LENGTH': '83',
'CONTENT_TYPE': 'application/x-www-form-urlencoded',
'DOCUMENT_ROOT': '/data/web/vhosts/portal.g-node.org',
'GATEWAY_INTERFACE': 'CGI/1.1',
'HTTPS': '1',
'HTTP_ACCEPT': 'text/html,application/xhtml+xml,application/xml;q=0.9,/;q=0.8',
'HTTP_ACCEPT_CHARSET': 'ISO-8859-1,utf-8;q=0.7,*;q=0.7',
'HTTP_ACCEPT_ENCODING': 'gzip,deflate',
'HTTP_ACCEPT_LANGUAGE': 'en-us,en;q=0.5',
'HTTP_CONNECTION': 'keep-alive',
'HTTP_COOKIE': 'sessionid=d63df70e77b7b6dcbe4287c82f72ba1c',
'HTTP_HOST': 'portal.g-node.org',
'HTTP_KEEP_ALIVE': '115',
'HTTP_REFERER': 'https://portal.g-node.org/data/account/password_change/',
'HTTP_USER_AGENT': 'Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2.25) Gecko/20111212 Firefox/3.6.25 ( .NET CLR 3.5.30729; .NET4.0E)',
'PATH': '/usr/local/bin:/usr/bin:/bin',
'PATH_INFO': u'/account/password_change/',
'PATH_TRANSLATED': '/data/web/vhosts/portal.g-node.org/account/password_change/',
'QUERY_STRING': '',
'REMOTE_ADDR': '141.5.20.126',
'REMOTE_PORT': '6705',
'REQUEST_METHOD': 'POST',
'REQUEST_URI': '/data/account/password_change/',
'SCRIPT_FILENAME': '/opt/g-node-portal/apps/g-node-portal/apache/django.wsgi',
'SCRIPT_NAME': u'/data',
'SCRIPT_URI': 'https://portal.g-node.org/data/account/password_change/',
'SCRIPT_URL': '/data/account/password_change/',
'SERVER_ADDR': '141.84.44.72',
'SERVER_ADMIN': '[email protected]',
'SERVER_NAME': 'portal.g-node.org',
'SERVER_PORT': '443',
'SERVER_PROTOCOL': 'HTTP/1.1',
'SERVER_SIGNATURE': '

I think it should be impossible for an benchmark creator to take this benchmark offline again or change the underlying datafiles once someone else has used, e.i. submitted an evaluation, for that benchmark.

That could e.g. lead to someone really using the website for his evaluations and suddenly the benchmark disappears, or worse changes without him noticing.

please close immediately

I just received an invitation from the G-Node portal since someone added me as collaborator to one of his benchmarks. The link in the mail to accept that request is broken and leads here:
https://portal.g-node.org/account/login/?next=/data/invitations/

in one case the ground truth file failed the test. I want to update
only that ground truth file and not upload the raw data again, since
the raw data was fine. I should not have to update all entries (also
the details and name) and both files again

django-based query methods should work not only for objects attributes but for special filters also.

Add record:

• what is a record? didnt we call that differently? I recall "benchmark", "subbenchmark" and "data file", "benchmark file" or "data file pair".
• upload of a record can take long but has no status bar
• there must be a description of the file formats for the uploadable files
• the "view log" text is not clickable (only mouse over).
• the log text could be improved (e.g. after uploading a wrong gt
  file): "looking at groundtruth file version with uid: 62 error during
  record check: invalid literal for int() with base 10:
  '\xcf\xfc\x8c\xa63x(M\xf2y\xf1'"
• there is a line break between the "x" icon for failed uploads and
  the "view log" text

Low priority:

• option to upload several records at the same
  time. the 8+4+4+4 files of the Quiroga benchmark will be a pain
  otherwise

Preliminary by using a django dev version, prefetch_related.

i can see and click on Andreys evaluation even though i dont have
the right to see it (i get an error page)

After reading a private message on the portal it still was shown as unread.

Change all exceptions to local ones (define new classes) so not to loose real bugs.

Validate e-tags while updating an object.

If you create a new benchmark and you create records belonging to that benchmark. Suppose one of these records fails the validation. The user should not be able to make this "broken" benchmark public.

my first evaluation on the Quiroga2004 benchmark, record
Easy1_noise025 shows not the algorithm title although it was set (you
see it, if you click on "Profile for ffranke", the name appears)

Put them in JSON and transfer as one key-value? Any other options?

Description:
On my profile page is
a) a text box not where it should be
b) a broken link

How to reproduce bug:
a) Login into the portal under https://portal.g-node.org/
b) click on "Profile", the "your profile" tab is active. (i am now here https://portal.g-node.org/data/profiles/profile/ffranke/)
first bug:
The Text "Name: Felix Franke Location: ETH Zürich, D-BSSE Website: http://www.bsse.ethz.ch/bel/people/frankef" overlaps with the BCCN logo
c) click on the "profiles" link under "your friends" (I have none). The link directs to: https://portal.g-node.org/profiles/ which is broken

set parameter for record (e.g. noise level),
used for generating the "between evaluations summary"-plots

parameter should have a name (one per created benchmark)

no button to go back to benchmark overview

a benchmark can be made "active" although records failed the unit test

We need dynamic pages for the following issues and have to discuss on a) what we actually want, b) which technology to use for that.

1. Documentation for the user. We need to put text/figures/explanations on the website on specific items/buttons/figures to explain the user how to use the website and what the output means.
2. We might want a dynamic part for evaluations an benchmarks where users can leave comments
3. The creator of the benchmark might want to add details to his benchmark and add information about what the benchmark means

Maybe one content management system can handle all three, or two of them?

change ordering of records within a benchmark should be possible

There should be an opportunity to share objects on all levels (segment, signal, unit etc.) without duplicating objects or changing the dataset structure.

The website needs more explanations for users of the items it depicts.

Example: When looking on the result page of an evaluation, the table containing the error values needs to be understandable.

Problem: These explanations can be very long and might contain pictures. Furthermore, it would be desirable if someone (Felix) could modify them without having to bother the developers.

Task: Decide either on an content management system, external wiki or other solution that will be used to manage the explanations.

It seems that github pages (http://pages.github.com/ ) service doesn't work as expected for organizations. I've succeeded to create a page for G-Node (g-node.github.com) but it doesn't work for repos (however, always says that page build was successful). If someone was different results pls let me know.

My example pages as a gh-pages branch in g-node-portal repo:
https://github.com/G-Node/g-node-portal/tree/gh-pages

Thus for the moment I propose to use single repo for documenting all projects, if needed. This repo is called g-node.github.com, everything that goes to master branch there is being automatically built to g-node.github.com.

Andrey

Add timestamp to file operations log. And maybe make it as a dict, not just a string, adding task ids.

Lets say a benchmark has 8 records. A user sorts all those records and gets 8 .gdf files. When uploading all those files he needs to do so for every file individually. That is boring and takes time.

Solution:
There are several options to solve that problem. Some are discussed below as "too much work". The simplest is probably:
Like in the prototype, when clicking on upload evaluation for a record, show an edit field for every records and allow that the user adds a file to all those edit fields.

Other solutions (probably too much work to implement):

• Allow for uploads of .zips containing many files. The files must be named in a way that the website understands to which record to put them, e.g. folders for benchmarks, filenames for records.
• Allow for multiple file selection when opening the upload dialog. Now the website has to derive from the filenames for which record the files are.

When uploading a gdf file that contains a sorting and this file is in some way corrupt, e.g.

• it contains only one column,
• it contains floats instead of integers,
• it contains binary data
• it is too big

the error should be reported to the user and the explanation of the correct data format should be shown. Maybe even with a link to the Matlab function that stores gdfs :

X = [001 101
002 102
001 215
001 469];
filename = 'dummy.gdf';
delimiter = ' ';
X = X';
fhandle = fopen(fileName, 'w+');
fprintf(fhandle, ['%05d' delimiter '%d\n'], X(:));
fclose(fhandle);

On the detail page of an evaluation there should be a button that allows downloading the results.

I would suggest the easiest way possible, but I dont know what that is. Would it be possible to build a zip file that contains

• a) a H5file/matlab file containing the result table
• b) a text file containing the result table as depicted on the website
• c) the figures from the evaluation page
• d) a readme file specifying what that zip contains, which benchmark was used and which evaluation that was

The rendering of the friends in the top-right friends box is off (see screenshot: http://people.gnome.org/~gicmo/portal-bug-friends.png ).

if i click on andreys benchmark with the name "FelixFranke" i see
the link to "view evaluations (1)". if i click it i see all
evaluations, also the ones on my benchmark "Quiroga2004"

PROPOSITION: The disclaimer page should present the current and correct affiliations (GNode, INCF, TU, BEL). Currently we have a copy of the page that was in the prototype. Could someone please come up with a current version or replacement.

EXTENDS: http://141.84.42.103:31337/spike_evaluation/disclaimer/

MEDIA: http://www.g-node.org/overview/contacts/overview/contacts/imprint

When an evaluation is uploaded, the user has to specify how he created the sorting. So far we called that "algorithm". However, I think it would be useful to rename that field since there are also manual sorters around.

Therefore, rename the field to "method or algorithm".

Although I click "Remember me" all the time during login as soon as I navigate away from the portal it forgets that I am logged and will ask for the login credentials again.

When editing a benchmark record, you can e.g. change the parameter for noise standard deviation. But if you click on "edit" nothing happens. The benchmark is in fact successfully edited, but the user does not see that.

preliminary using built-in django multiple-objects-update mechanism.

Tag- and untag- objects recursively.