Software Problems That Motivate Web Searches - Finding Source Code on the Web for Remix and Reuse - page 255

Fig. 13.1 Problems that motivate web searches
of the duration of each kind of search. A Search Session represents a period of
continuous Web usage to fill a single information need performed in the same day. 1
Figure 13.2 shows a box and whisker plot for the length of the searches by the
type of search performed. Each box in the graph shows the range of 50% of the data
and the black dot shows the median. The whiskers show the 25% of searches that
took longer or the shorter. Triangles show the outliers. (Three additional outliers
have been omitted.)
Here, we include a description for each type of software problem identified for
opportunistic searches. Descriptions are accompanied by quotations from develop-
ers that exemplify the searches our participants performed in field studies. We use
pseudonyms to identify our participants.
1 Some definitions of sessions used by other researches indicate that Web usage must be continuous
with no breaks longer than 5 min [ 11 ], 6 min [ 3 ], or 25.5 min [ 4 ]. However, we decided not to
include a cutoff in the longest time a break could take in the same day. Our decision was based
on the fact that we observed that developers had interruptions longer than 25.5 min while using
the Web and when they came back they continued reviewing information on the Web or refining
queries to fill the same information need. For that reason, the only time constraint we consider is
that Web usage should happen the same day. Instead, we put more emphasis to the intention of the
search. If Web usage was done to meet the same information need, we consider all those intervals
as part of the same Search Session. The three longest breaks that we observed in our field sites were
of 3 h 25 min 46 s, 1 h 34 min 19 s, and 1 h 16 min and 18 s. In these cases, developers interrupted
their search to chat with co-workers, answer calls from customers, code, write documentation, and
have personal breaks. When they used the Web again, they read the information they found before
and in some cases they also copied and pasted lines of source code and used them even though
more than 3 h passed since their first query.