Estudo empírico de análise da compatibilidade de aplicações Android com diferentes versões da API da plataforma

Abstract

Android is currently the most popular platformfor the development of mobile applications, representing more than 80% of the operating systems market for mobile devices. This causes demands for application customizations to handle different devices such as screen size, processing power and available memory, languages, and specific user needs. Twenty-three new versions of Android platformhave been released since its first release. In order to enable the successful execution of applications on different devices, it is essential to support multiple versions of the Application Programming Interface (API). This dissertation aims to analyze, characterize and compare techniques used by Android applications to support multiple versions of the API. In particular, the work seeks: (i) to identify the used techniques to support multiple versions of the Android API in the literature; (ii) to analyze real applications to quantify the usage of these techniques; and (iii) to compare the characteristics and consequences of using such techniques. An empirical study, in which 25 popular Android apps were analyzed, was conducted to achieve this goal. The results of the study show that there are three techniques to support multiple versions of the API: i) compatibility package, that adrresses API coarse granularity variabilities involving a set of classes; ii) re-implementation of resource used for specific situations and coarse granularity at class level or when resource is not available in compatibility package; and iii)explicit use of the new API that allows implementing fine grained variabilities of the API that involves calling of specific methods. Through the analysis of 25 applications, we have identified that compatibility package was used by 23 applications, re-implementation of resource was used by 14 applications and the explicit usage of the new API was used by 22 applications. The API fragments contains the most common elements among those released in higher versions of the platformthat are used by applications during their evolution, and it is referenced by 68% of them. In general, applications could increase their potential market with adaptations of, on average, 15 code snippets. On the other hand, application developers have been worried about how avoiding dead code based on platform API. In the analysis of 7 applications, 4 of them contained dead code, but it did not represent more than 0.1% of total code.