Citation Context

...ys obliviousness. A declarative approach to describe error recovery strategies is by using error recovery rules [2, 27]. Using automated analysis of a grammar it is even possible to derive such rules =-=[12, 17, 32]-=-. Derivation of rules is based on recognizing typical programming language patterns, such as scope structures or string literals, for which recovery rules can be formulated. For such an analysis to wo...

Citation Context

...ally changed one. For example, a parse error (e.g., “’;’ expected”) during incremental parsing may result in a corrective measure being taken (e.g., “insert ’;’”) in order for the parsing to continue =-=[44]-=-. Final BX pushes this change back to the input, so that the error can be corrected at the source, most likely by some kind of user interaction, such as Eclipse’s Quick Fix feature. 5 Case study In or...

Citation Context

...ements, the size of the generated symbolic output, the size of JS code, and the number of fixed errors are listed in Table 1. 6.1 Practicality and Accuracy All call-graph computations completed within a few seconds (< 8 seconds on average, < 12 seconds in the worst case). Since this performance is acceptable for executing analyses in the background of an IDE, we did not perform additional rigorous performance measurements. Currently, we need to fix all HTML syntax errors before we can build call graphs, which is a positive side effect, but also a laborious endeavor. An error-recovering parser [17] or automated repair tools [46] could help with this step. Overall, our tool is easy to set up for a new project by pointing it to the main file(s). Precision. In Table 2, we list the number of call-graph edges detected in all subject systems. We manually checked the created call graphs for correctness. In small call graphs (< 100 edges), we inspected all call-graph edges; in large call graphs, we randomly sampled 50 edges each. We consider a call-graph edge as correct when it connects two nodes tracked to PHP string literals, and the nodes are actually in a call relationship. All 604 checked ...

Citation Context

...s generalized-LR (JSGLR) parser is relatively efficient, ensuring good runtime performance and allowing for interactive use of the Spoofax testing language. With support for error recovery techniques =-=[9, 23]-=-, JSGLR also ensures that a valid abstract syntax tree is produced for providing editor services in case a test module is syntax incorrect (as seen in Figure 6 (a)). There are still opportunities for ...