Mossum is a tool for summarizing results from Stanford's Moss. The tool generates a graph for (multiple) results from Moss, which can help in identifying groups of students that have shared solutions.
The tool can also generate a report, which shows which solutions are similar between all pairs of students. When submitting multiple parts of an assignment to Moss, this can help in identifying which students have multiple similar solutions.
The script can be installed via pip:
pip3 install git+https://github.com/hjalti/mossum@master
Please note that this project is only developed and tested using Python 3.
After installation, the script can be called as follows
mossum MossURL1 MossURL2 ... MossURLn
If no URLs are specified as parameters, the URLs are read from the
standard input. In this example, mossum generates n PNG images, that
show the relationship between the submissions in each result page. The name of
each image is obtained from the comment used when submitting to Moss (the -c
flag).
By default mossum only shows a link between two submissions if the match Moss finds
- accounts for more than 90% of either submission, and
- accounts for more than one line of either submission.
The percentage and number of lines can be changed using the --min-percent
(-p) option and --min-lines (-l) option, respectively.
For example, when called with
mossum -p 95 -l 30 MossURL1 MossURL2 ... MossURLn
mossum only shows a link between submission if either match accounts for more than 95% of the submission or more than 30 lines are matched.
When submitting to Moss, the whole path of all files is shown. When summarizing, we can extract relevant information from the path. This can also be useful if assignments are submitted in multiple files. In which case, we can extract relevant information from the path (e.g., student ID), when summarizing.
The transformation option takes a regular expression as a parameter. If the regular expression contains groups, the transformed submission names are formed from the strings matched inside all groups, joined by an underscore. If the regular expression contains no groups, only the part of path that matches the regular expression is used as the transformed name.
For example, if the submissions are stored in the following directory structure
<assignment_name>/<assignment_id>/<student_id>/<assignment_file>
for example,
assignment1/1337/fred24/a1_part1.cpp
assignment1/1337/fred24/a1_part2.cpp
assignment1/1337/fred24/a1_part3.cpp
assignment1/1337/sarah37/a1_p1.cpp
assignment1/1337/sarah37/a1_part2.cpp
assignment1/1337/sarah37/a13.cpp
the student ID can be extracted using the regular expression .*/(.+)/.* and
if we want the name to contain the assignment name and the student ID, we can
use the regular expression (.*)/.*/(.*)/.*.
By extracting only the students' IDs from the path, the output might look something like this.
The summarization can be anonymized using the --anonymize or -a flag. When
using this, all names are substituted with a random name. This is useful for
frightening students.
When assignments are split into parts and each part is submitted to Moss
separately, it can be hard to get an overview of which students have many
submissions in common. If the --report or -r flag is specified, mossum
generates a report. The report generated shows which matches each pair of
students have in common. The report is sorted, such that the pair of students
with the most matches in common come first.
Calling
mossum -r -t "*./(.+)/.*" MossURL1 MossURL2 ... MossURLn
would yield a report with entries similar to
fred24 and sarah37:
part1: http://moss.stanford.edu/results/???/match??.html
part2: http://moss.stanford.edu/results/???/match??.html
part3: http://moss.stanford.edu/results/???/match??.html
The reports are stored in <filename>.txt, where <filename> is either
specified by the --output or -o flag or obtained from the name of the
assignments (extracted from the Moss pages).
The format of the image generated by mossum can be changed using the --format
or -f option. Mossum uses Graphviz to generate graphs. All output formats
supported by Graphviz are therefore supported by mossum. See Graphviz output
formats.
Edges are labeled, by default, with the percentage of code matched (the higher
percentage is used) and the number of lines matched. Additionally, hyperlinks
to the Moss page of the match are added to edge labels. However, hyperlinks
only work if the specified format is svg or xlib.
For example, calling
mossum -f svg -t "*./(.+)/.*" -o assignment MossURL1 MossURL2 ... MossURLn
will generate a SVG image file. If this file is opened in a browser, the matches found by Moss, which represent each edge, can be opened by clicking the label displayed on each edge.
If you are using X, you can specify the format xlib. In this case mossum
displays the graph in an Xlib window, and the graph is therefore not saved to
a file.
Edge labels can be hidden with the --hide-labels flag.
It is possible to to add or remove a name from the collection of assignments using filters. There are four filter options, all of which take a list of names as arguments. The following filters are available.
-
--filter N1 N2 ... Nn: Only show connections between the specified names. No other connections are shown. -
--filteri N1 N2 ... Nn: Only show connections including at least one of specified names. No other connections are shown. -
--filterx N1 N2 ... Nn: Do not show connections between the specified names. All other connections are shown. -
--filterxi N1 N2 ... Nn: Do not show connections where one of the specified names is involved. All other connections are shown.
It is also possible to filter out nodes that are connected by more than one
edge using the --min-edges option. This is only applicable to merged results.
This can be useful, e.g., in assignments with multiple problems. Submissions to
each problem can then submitted to Moss separately. Merging these results can
lead to a lot of noise. For example, like this.
Applying --min-edges 2 highlights students sharing more than one solution.
