From work to word: How do software developers describe their work?

Abstract

Developers take notes about their work sessions, either to remember the work status and share it with collaborators, or because employers explicitly require this for project management matters. We report on an exploratory study which aims at understanding how software developers describe their work. We analyzed more than 750,000 work descriptions of about 2,000 professionals taken over 8 years in three settings. We observed several similarities in the content and time meta-data of work descriptions. Most frequent terms, such as top-30 performed activities, are used consistently. Particular templates such as "ACTION concerning ARTIFACT because of CAUSE" occur frequently. Developers described sessions that last 30-120 min. 4-16 times a day. Maintaining diaries seems to consume between 3-6% of the total work time, and in 10% of the sessions, developers did not describe their work in sufficient detail. We argue that our results make the first step towards automatically generating work diaries for software developers.

The MSR Cookbook: Mining a decade of research

“…Alternatively, log weight functions are sometimes used to lessen the effect of a single term being repeated often in a corpus [24]. Depending on the context, you may then need to create various indices: a standard term index including all terms, a stop-index which excludes stop words, a stem-index with stemming applied, and a synonymindex [24] that aggregates synonyms for semantic analysis [24], [39]. Depending on the technique used and the analysis applied, normalization and smoothing may also be applied [49], [57].…”

A Survey on Mining Software Repositories

“…Decision tree has been used to predict developers' contribution in [145]. In [65], SVM has been applied for the bug triage and in [147], association rule mining has refactoring [72], [73] API-change [74], [75], [77], [80], [81] change patterns [83]- [88], [90], [160] team-activity developer's contribution [55], [91], [93], [94], [154] experties of developers [96], [97], [149] tool support [98], [99], [128], [151] helpful information [100] comprehension visualization [101], [102], [156] identifiers [104], [105], [153] recording operations [106] validation metrics [45], [107], [157] tool [108] clones [109]- [112], [150], [159] bug [113], [114] development& evolution development [118]- [120] evolution …”

“…Maalej and Happel explored the way software developers describe their jobs [118]. For eight years, they analyzed 75,000 work descriptions of 2,000 professionals and found that there are similarities between metadata of contents and time in the description; the typical pattern is "ACTION concerning ARTIFACT because of CAUSE" [118].…”

“…For eight years, they analyzed 75,000 work descriptions of 2,000 professionals and found that there are similarities between metadata of contents and time in the description; the typical pattern is "ACTION concerning ARTIFACT because of CAUSE" [118]. Developers did not describe their job in detail.…”

Assisting engineers in switching artifacts by using task semantic and interaction history

“…A case study done by Zou [21] found that 8 code files are read and 6 are changed during a task. As engineers typically spend only 50% of their time for code creation [13], and as Zou only observed code files inside IDEs, it is most likely that the overall number of used artifacts is even higher. Multiple tasks in conjunction with multiple artifacts per task increase the amount of artifacts the engineer has to deal with in parallel.…”