CVS log for loncom/interface/statistics/lonproblemanalysis.pm

Up to [LON-CAPA] / loncom / interface / statistics

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

- For 2.11
  Backport 1.147

- Load CSS file(s) for resource when using (a) Student Submission Reports
  (only one problem selected), or and (b)Detailed Problem Analysis when
  "Show problem" is checked, or when using Submission Time Plots.

- For 2.11
  - Backport 1.145, 1.146.

- Remove unnecessary white space in <select> and <option>.

- Fix typo (div)
- XHTML

- For 2.11
  - Backport  1.144.

- Internationalization and localization of labels in statistical graphs
- Improved quantity in result phrases about correct and incorrect submissions
- Added missing phrases and translations to de.pm

- For 2.11
  - Backport 1.143.

- Localization. Separate sentences receive separate &mt() calls to facilitate
  reuse.

- For 2.11.
  -reverse changes in 1.140 - "3.0" grading interface.

- changed headlines to normal text with info/error/warning style where necessary
- changed tables to data tables to unify UI

Bug 6455 - Progress Indicator now jQuery. Still need to test across browsers.

Add statistics and reports under "grading"

Optimized status bar localization:
Overhauled status text creation to allow proper translation.
    - Translate whole sentence instead of unhandy phrase fragments concatenation.
    - Simplified way of time text creation by using quant.
    - Removed code for not anymore supported old browsers (bug 1382).
    - Indirectly corrects mt() call with incorrect parameter (_2 -> _1)

Consistent screen ergonomics:
- First checkbox, then option text
- Removed special styles

Moved output options checkboxes from translation phrases to script code.
This ensures that
- the checkboxes are always properly displayed, independent from the translation phrases
- the place of the checkboxes is given by the script code to have a consistent screen layout, independent from the translation of the corresponding text

added quant-calls in the source and edited them in de.pm, because they were not used elsewhere

Added missing ";" to all " "

- Unbalanced HTML tags.

Replaced
  <nobr>...</nobr>
by
  <span class="LC_nobreak">...</span>

- lonsearchcat.pm: Added missing start tag for nobreak area
- lonsurveyreports.pm: Added missing end tag for nobreak area

Optimized option selection tables for statistics:
- Use LON-CAPA standard tables
- Added missing &mt() calls
- XHTML conform "checked" and "selected" usage
- Added info/warning/error style to info/warning/error messages
- Added text "Options" to headline of options column
- Added horizontal line before and after problem output for better separation of problem content and rest of screen

Additional changes:
lonproblemanalysis.pm:
    - Now uses LON-CAPA standard table for foil value table
    - Moved status bar from inside the options table to under it
lonproblemstatistics.pm:
    - Moved Graph option from under the table to inside option table
lonstudentsubmissions.pm:
    - Moved output_selector from headline to extra column inside option table

Optimized status display:
- Set text box for display purposes only to readonly
- Use standard nobreak style
- Same paragraph structure for similar pages

- updating to work (simplely) with multiple correct answers

- bug5095 - handle case of no students and no spread in provided answers
            more correctly (rather than infinite loop)

Bug 4954.  Filter title changed from "Enrollment Status" to "Access Status".  Access Status selections in lonhtmlcommon set to:
Currently Has Access
Will Have Future Access
Previously Had Access
Any Access Status
See comment appended to bug 4954 for more information.
Documentation updated to include ability to selectively display students with future access.

Apparently omitted May 5th commits which added groups filter to STATS functionality.

&Apache::lonnet::unescape -> &unescape
&Apache::lonnet::escape -> &escape

- update statistics calls to bread crumbs

Fix #4609, at least in the most superficial sense; it no longer ISE's.

Pending implementation of the "Detailed Problem Analysis" output into
CSV and Excel, remove the selection box for Excel & CSV so we aren't
promising what we can't deliver.

- BUG#4445

Bug 2989: Need more submit buttons on problem selection pages.  Problem
analysis and submission time analysis now have submit buttons after each
sequence/folder.

- ENV -> env

Make sure we ahve a value (not undef) for the number of correct answers.
Fixed unfinished code in numerical response analysis.

lonstatistics.pm:Removed package variable @SelectedSections.
    Added &get_selected_sections
    Added &section_and_enrollment_description to provide a brief
        description of those students selected for statistics work
        (hopefully this descriptive line will make it in to bug reports).
Added &section_and_enrollment_description to all statistics pages
Replaced use of @SelectedSections with &get_selected_sections in statistics
    pages.

Removed debugging spew.
OR analysis: Set $correct to 0 if it's undefined so we output '0' instead
of ''.

Refactored to use navmaps and modified lonstatistics/lonstathelpers code.

lonstathelpers:Added &submission_report_form which provides a direct link
    to the student submission reports page from another statistics page.
    It will not work from any other component of LON-CAPA.
lonproblemanalysis:Added call to submission_report_form

Finishing off NRA.  Plots of percent difference and difference can be made.

So the last commit of code for &numerical_plot_percent was stupid for
plotting percents, but much better in &numerical_plot_differences, which
now works well.

half way through rewrite of numerical response analysis.
Added progress window to student answer computation.
&numerical_plot_percent is mostly done.
&numerical_classify_response gathers more statistics and passes them back
    in a seperate hash.
Minor code cleanups elsewhere.

Remove excel output of option response data.

Disable numerical response analysis entirely

re-enable numerical response analysis with units.

Fix bug with excel output of option response results which caused data to
be omitted for foils which contained spaces in their names.

Restrict analysis to implemented types.
OR_tries_analysis: No longer ISE when there are no concepts.

"finished" numerical response analysis - unit conversion does not work
reliably yet, so it is disabled.
lonstathelpers: now &get_problem_data returns the 'unit' for a response.
lonproblemanalysis: Added earlier detection of broken problem analysis
  (See bug 3669 for why this came up)
  radio_response_analysis: Make sure $concepts is at least an array ref
     when it's undef, just to avoid an ISE.

Modified option response analysis to label every column in the graphs by
passing xskip=1 to &Apache::loncommon::DrawBarGraph

Finished time analysis for numerical responses.  Things seem to work but I
am seeing strange results.

Numerical response analysis works for attempts, but not time analysis.
Shows histogram of correct answers each time, which is dumb.

Added count of students to radiobutton response time analysis.

RR_create_percent_selected_plot:
Fixed bug in prior commit where radiobutton submissions were plotted in
the column corresponding to the correct answer and not the submitted answer.

Removed some debugging code.

RR_create_percent_selected_plot: Fixed bug where if a foil could be true for
   some students but not all students the bar for the foil would appear
   green (indicating true) for everyone.  Now keep track of those for whom
   the foil is true seperately than those for whom the foil is false and
   produce a stacked bar graph (green & red = true & false).

radiobutton response analysis: Added %true_foils to keep track of which
   foils are true.  Modified RR_create_stacked_selection_plot to use
   this hash to ensure all correct foils show in the stacked plot.

Radiobutton response works for concepts, computed answers, and the much
more common incarnations.
Removed some dead code.
More to be done.

Modifications to option response analysis.
Tries & time based analysis:
    use 'statements' instead of 'foils' on plots.
    list number of students attempting, number of submissions plotted,
    number of correct submissions, and number of incorrect submissions.
time based analysis:
    Fixed a bug where the concept and foil graphs appeared to have a different
    number of data points to plot by consolidating the answer post-processing
    code into &OR_time_process_data, which needed to happen anyways.

Bug 3571: Option response analysis option key colors did not print because
most web browsers do not print table background colors.  Using
background colored asterisks instead of   fixes this.

Radio button response analysis over tries is implemented.  Check in to save
my work.

Fix bug wherein option response tries analysis always showed the same
concept graph.

OR tries analysis now checks for missing data and gives a friendly
message instead of showing a bunch of empty graphs.

Added 'show problem' checkbox.  Time analysis would always show a plot, now
it does not show a plot if there is no data.

Interface & localization changes

Remove help reference to nonexistant help topic.  "Number of Plots" is pretty
damn self explainatory to me.

Interface simplification: got rid of "analyze as" nonsense.  We try to
produce three plots for each attempt or time interval: concept, foil correct,
and foil incorrect.  If there is only one concept the concept plot is omitted.
Some Style Police(tm) induced changes.  Titles of plots are more consistent
(since they appear on the same page now).

Removing access to numerical and radio response analysis for 1.2 release.

No need to tell the user there are no students in the selected section
before they have chosen a problem to analyze.

Unified cache handling among statistics routines.  All use inlined progress
dialog.
&lonstatistics:&Gather_Full_Student_data now takes $formname and $inputname
    for inlined progress dialog.
&lonstathelpers: All calls to lonstatistics::Sequences_with_Assess now pass
    parameter 'all' in for mode.  Added &manage_caches which takes care of
    updating student data at the right times.  "right times" may need some
    work.  &GetStudentAnswers now takes $formname, $inputname as parameters.
&lonproblemanalysis, &lonproblemstatistics, and &lonsubmissiontimeanalysis:
    Use lonstathelpers::manage_caches and inline progress dialog.
&lonstudentsubmissions: Uses lonstathelpers::manage_caches and inlines prgwin
    Uses inlined prgwin to display further status data.

- missed and encode fixup

- trying to be UTF8 transparent

loncoursedata: reworked &get_response_data to take section and enrollment
as inputs instead of a list of students.
lonproblemanalysis: Rewrote calls to get_response_data to provide new inputs.
Modified numerical response analysis to analyzse an entire problem, not just
one response.  This will probably need to be changed in the future.
Added &text, which produces <drawimage> text elements.

Moved &GetStudentAnswers from lonstudentsubmissions and lonproblemanalysis
to lonstathelpers.

lonstathelpers: added local caching of computed answers.
lonproblemanalysis and lonstudentsubmissions: modified to begin using the
caching of computed answers.

Check for aborted connections when compiling results for students.

Reworked analysis of numerical responses.  Uses <drawimage> and &xmlparse
to create a chart of answers.

OR_Tries_Foil_Analysis: Changed plot titles to 'Submission X (N=Y)' where
X is the attempt number and Y is the number of students.

OR_Tries_Foil_Analysis: Added a blank column between concepts on the plots.

&OR_Tries_Foil_Analysis: Variable name cleanup,  Use $Concepts instead
of $Foils for looping.

No longer translate before calling lonhtmlcommon::breadcrumbs
and lonhtmlcommong::add_breadcrumb.

Translate text passed in to &Apache::lonhtmlcommon::breadcrumbs and
&Apache::lonhtmlcommon::add_breadcrumb.

Modified to use &Apache::htmlcommon breadcrumb routines.

Added lonstudentsubmissions.pm
Moved code from lonproblemanalysis to lonstathelpers and lonstudentsubmissions.
lonstudentsubmissions can produce excel spreadsheets of student submissions
for numerical, radio, option, click on image, string, and formula responses.

Modified &prepare_excel_output to include the 'awarded', 'weight', and
'score' columns in the Excel worksheet it produces.

First working pass at excel output of student submissions and the correct
answer for a response.  More needs to be done.

Some progress on numerical problem analysis.  Makes a plot that is nearly
useless.

Fix bug with non-numeric concepts screwing up option response analysis.

Added the very beginnings of detailed analysis for numerical problems.
Currently just prints a list of the correct answers by student.

No need to encode text that must work when output as html.

Added \@Labels to DrawBarGraph inputs (labels on the x-axis).

Bug 2675: html entities not handled correctly in option response analysis
display of concepts and options.  Can go in 1.1.2 without issues.

Changed the title of the detailed problem analysis page to read
"Detailed Problem Analysis" insead of "Option Response Analysis".

Moved &ProblemSelector from lonproblemanalysis and lonsubmissiontimeanalysis
to lonstathelpers.pm.  Reworked &ProblemSelector to work with either problem
parts or responses.  Added previous and next buttons to submission time
analysis page.

Added entry for lonstathelpers.pm in loncapafiles.
Modified lonsubmissiontimeanalysis and lonproblemanalysis to use
lonstathelpers.

Finish renaming option response routines.

1. Removed debugging code.
2. Unescape the foilid.  This fixes a bug with spaces in foilids causing
data to be missing in the plots.

1. Changed most if not all option response related subroutines to begin
with 'OR'.
2. Placed code to list the type of the problem before it's chosen.  This will
be removed (or reworked) soon anyways.

Implement basic radio response analysis.

Removed debugging information.

Bug 2556: Previous and Next problem buttons on detailed problem analysis.

Use 'problem' instead of 'resource' when describing what we are
analyzing.

Added help file and help link for problem analysis.

Added help links for "Analyze Over" and "Analyze as".  Please include in
1.1.

&ProblemSelector(): fallback for resources without titles.
The wonders of cut and paste coding.

Use the new &Apache::loncommon::store_course_settings and
restore_course_settings subroutines to remember form settings.
lonstatistics.pm was changed to not parse the student enrollment status
form elements before calling the sub-module routines which present the
interface.  Now each of the sub-modules (statistics/*) must call
&Apache::lonstatistics::PrepareClasslist(); at the right time.

Old changes that need to be committed.  Initial stab at radioresponse analysis
is 'POD'ed out.  It works.  Trust me.

Works with new version of loncoursedata.pm.  Radio response analysis is
not fully functional, although you can select those resources.

Reorganization of code, some subroutines renamed.  Goal is to make the
implementation of new types of problem analysis easier.

Fix idiocy with time analysis.

No longer perform analysis when excel output is requested.
Cosmetic changes, nothing shocking.

Excel output for optionresponse analysis.

loncoursedata.pm:
Added &get_student_data.
Modified return values of get_optionresionse_data it now logs errors and
(on error) no longer returns a reference to useless data.

lonproblemanalysis.pm:
Added 'Produce Excel Data Sheet' button.  Wording will probably change, as
will page which presents the spreadsheet for download.
Added a bunch of routines associated with excel output, including two
routines 'calc_serial' and '_adjustment' copied from Andrew Benham's
datecalc1.pl, part of the Spreadsheet::WriteExcel CPAN module (GPL).
Read through some of the comments on these to learn a little about date
handling in Excel.

Many changes, most to internal code structure.  Hopefully future changes
will not require such monsterous rewrites.
Foils analysis now shows two graphs, one for the correct choice on the
foils and the other for the incorrect submissions.
Time analysis no longer has editable plot titles.
Added a few helper functions get_time_from_row, get_tries_from_row, and
Process_Row.

Removed code which strips leading and tailing whitespace from problem data
because Guy told me to.  Minor changes otherwise.

*** empty log message ***

Monstrous rewrite to, on foil analysis, output graphs showing how the
students got the foil incorrect.  As a side effect there will no longer
be any big green bars on the top of the bar chart in concept analysis
mode.  The case of one concept is handled more cleanly.
Only one column of graphs in foil analysis mode as the key needs to be
repeated.
Skip 'MISSING_ANSWER' graded submissions because they will really screw it
all up.
The major functionality to include is implemented, with the exception of
Excel output of the basic data.
Assumes no options = '_total' or '_incorrect'.
Known issues:
The 'submission' and 'submissiongrading' parameter thingys do not appear to
allow leading/trailing spaces, so these are zealously removed.  May be an
erronous assumption.
A warning is thrown: "Possible attempt to put comments in qw() list".  I need
to turn off warnings around the "offending" line of code.
HTML::Entities::decode does not decode %20 into ' ' so I have to do it myself.
The foil analysis plots can look pretty vomitous, especially for the
Atwood problem in the PHY183 Spring 2003 course.
The colors selected may not be useful to color blind people.

Trivial formatting change.  Nothing to see here.  Move along.

Concept analysis is now implemented for Times as well as Tries.
Time plots now have user-specifiable titles.
Removed old &logthis calls and unused code.

Concept analysis work for tries.  Much reworking of internals was required.
More to come.

Minor code cleanup.  Additionally, most humans begin enumerating with 1, not 0,
so the first graph is now called 'plot 1' not 'plot 0'.

Added date specification ability to time analysis.
Makes the interface seem really ugly.  Advice would be appreciated.

First attempt at 'time' based analysis - actually, just taking a fixed
percent of the responses and graphing them.
Added 'Update Student Data' button because, when viewing a 2100 student
course it takes a while to even check the status of the students data.
The data is checked the first time a problem is analyzed but only on
request after that.
Added &build_foil_index as every analysis will have to have an index
of the foils.
Some comments and code reorganization.

Put those headers on the table and not in some random array....

Major interface changes.  Still only do tries.  Do not print problem
selection dialogs when we are already looking at a problem, added a
button to get to a page where you can select a problem.  Let the user
select the number of graphs to display.  Hope to add Excel & csv data
file generation too, but not today.

Fix '</form>' in rendered problem.

Fix typo.

Many changes, some large, some small.
Added 'tries' selection box, although it doesn't work yet and this
functionality will likely be changed signifigantly in the future.
Use loncommon::DrawGraph instead of our own DrawGraph.
Added proper handling of selected students, instead of always running
analysis on all students in the cache tables.

Further internationalization of lonproblemanalysis.
Added count of students to the problem analysis graphs - please take a look
at your favorite class and let me know what you think of the interface.

Cleanups and added error checking to the entire module.  More is probably
needed.

Works for analysis of simple option response problems currently.  No
implementation of concept groups and no time-based graphs, just 1-3 tries
analysis.  I'm still developing the modules structure so it's a bit of a mess.

It's probably unneccessary to have people translate html tags.

Gutted before rewrite.  Committing for coments and safety.  Does nothing
useful except: displays a list of option response problems and lets you
select one.  The selected one is then rendered at the top of the page
(with relative links broken).

- ssi_body not ssi (yes Iknow I can't test this,but it feels right)

- converted lonspreadsheet to use new mechanism
- removed debugging sleep
- typo in lonproblemanalysis

- adding timing info to the popwin, and adding new functionality,
- converted stats to use new interface

lonstatistics.pm:
   POD cleanups
   @SelectedSections is set when we read the classlist.
   Added variables $top_map, @Sequences, and @Assessments.  These are
      initialized by &PrepareCourseData($r).
   Added &PrepareCourseData($r) which is essentially a wrapper for
      &Apache::loncoursedata::get_sequence_assessment_data().
   Added &MapSelect() to output a <select> box for sequences.  Not tested.
   Added &SectionSelect(..) to output a <select> box for sections.  Tested.
   Added &DisplayClasslist($r) which displays a table of the current classlist
      that is sortable by each column.  Does not bother with 'update time' as
      this should not be an issue for the user.  Tested.
   Added call to &PrepareCourseData by the handler.
lonpercentage.pm, lonproblemanalysis.pm, lonproblemstatistics.pm,
lonstudentassessment.pm: modified to call &Apache::lonstatistics::SectionSelect
   instead of the method in lonhtmlcommon.pm.

- using png now

tyr to fix bug 1039. I am not sure it has benn fixed.

Again cleaning the source.

Cleaning the source.

Fixed inconsistency in computing the Option Response Problems Analysis. This module "restore" a particular problem results for every students. Since this restore function is very fast, it is no need to cache the data any more. For example for 835 students in phy231 cousre it takes 33 seconds or in other word it takes for 100 students less than 4 second. We don't use "dump" here.

continue to fix bug 939
needs more work

continue to fix bug #939

Start to fix inconsitency in problem analysis and fix bug 939
It needs more work.

Fixed a bug in showing the problem Analyais table

Fixed the number not updating in the status window for update all students
button.  Also fixed some tieing to hash problems for lonproblemstatistics
and lonproblemanalysis.

Fixed a problem with statistics calculations, acts funny when testing near
zero.  Fixed sequence displays for problem statistics so that the sequences
are in the proper order.  Same with problem analysis.  Also removed
discussion column from statistics.

Added section selection to problem statistics and problem analysis.
Fixed some divide by zeros in lonproblemstatistics.
Broke up the table of option response problems and gave them the
heading of the sequence in which they lie.

No longer need to unescape values from ssi individually.

First, added unescaping of $key in lond dump command.
Next, I added a new way to download student course data.  There are now
two functions for storing data, DownloadStudentCourseData and
DownloadStudentCourseDataSeparate.  These two functions base their running
on input parameters.  The option parameters are whether or not to check
the date for downloading, whether or not to store all the dumped data or
extract out the data you want, whether or not to display a status window.
The extracting data parameter will be best utilized if someone adds in the
ability to send a list of what parameters are desired and perhaps some simple
commands to affect how that data is processed, like tries, sum would
sum record the sum of all the tries for a student.  This is just an idea.

Currently, I have all the statistics modules using the extract ability.
This slightly increases in download time, but drastically reduces cache
size.  Possible ideas include pushing the extract to the lond side with a
list of parameter/commands, or even downloading everything to a temp cache,
then extract the necessary data into the cache then removing the temp
cache.  There are lots of other possibilities, which can change the download
time, cache size, and other factors.  Now, only loncoursedata handles the
downloading of data to a hash.

lonstudentassessment was changed slightly to remove ' ' as a link if the
student actually hadn't attempted the problem.

lonproblemanalysis was updated for the new str2hash type functions.  There
are a couple of (cludges/fixes) for it.  Depending on whether or not the
str2hash type functions are changed, these may or may not need to be
updated.

lonproblemstatistics was drastically overhauled.  Most of the processing
was removed.  Now, it just does its few statistics functions and outputs
the table.  Currently, I broke the graph, discussion column, and
discriminant factor columns.  These will be fixed on the next commit soon.
There is also no caching done.  This will also be remedied soon.  The
problem that will need attention with caching is to know when to update
the statistics cached data when a student's course data is updated.

Lastly, I plan to add perhaps a toggle legend display button, another graph
button(percentage correct), a button to send the CSV format(not just display),
and add a toggle button for sorting within a sequence and sorting all
the problems.

Also, I changed the look and feel to be the same as the class list page.
Also, the displaying of sequence headers and child sequences are not
working.  This will be fixed, but thought will be put into how best to
make it look and have a similiar fill for all the table combinations.

Added in some flushes and changed where some of the modules print.  Also,
fixed download all so that if you stop it, and then select something
else it won't continue to download.

First, added the parenthesis thing to the GDBM stuff.  Fixed the interface
problem statistics so that the buttons work correctly.  How the data
is interpretted is not finished.

Fixed the problem gathering the correct data to be used for the analysis.
The problem stemmed mostly to the correct use of the result from ssi.

Broke lonstatistics up.  Now the bulk of the code that generate the
html pages for the different forms have there own module.  Some minor
modifications were necessary to accomplish this.

Also added default student selected on student assessment page is now
all students.  The full name links on the classlist page now take
you to the student assessment page with that student selected.  I
had forgotten to put the post data in the get_unprocessed_cgi
function call.