Speeding up Generalized PSR Parsers by Memoization Techniques

Abstract

Predictive shift-reduce (PSR) parsing for hyperedge replacement (HR) grammars is very efficient, but restricted to a subclass of unambiguous HR grammars. To overcome this restriction, we have recently extended PSR parsing to generalized PSR (GPSR) parsing along the lines of Tomita-style generalized LR parsing. Unfortunately, GPSR parsers turned out to be too inefficient without manual tuning. This paper proposes to use memoization techniques to speed up GPSR parsers without any need of manual tuning, and which has been realized within the graph parser distiller GRAPPA. We present running time measurements for some example languages; they show a significant speed up by some orders of magnitude when parsing valid graphs. But memoization techniques do not help when parsing invalid graphs or if all parses of an ambiguous input graph shall be determined.

Visual exploration of visual parser execution

“…PARVIS has already been used by the authors to visualize the execution of LALR parsers generated with CUP 2 and visual language parsers. Examples of the latter are Predictive Top-Down (PTD) [10], Predictive Shift-Reduce (PSR) [11,12], and Generalized PSR (GPSR) parsers [15,25] generated with GRAPPA 3 as well as pLR parsers [9]. As described in the previous section, we had to extend the parser runtime and write a reader script, so that the parser generates a log file during its execution, which translates the script into an event tree.…”

“…Here the concepts of hypergraphs, hypergraph grammars, and their parsers will be described only so far as to be able to understand their PARVIS instances. Details can be found, e.g., in [12,14,15,25]. Hyperedge replacement grammars (HRGs) are special hypergraph grammars and a direct extension of context-free grammars, such as those underlying LALR parsers.…”

“…The derivation process and derivation trees are defined similarly as for context-free grammars. We consider Generalized Predictive Shift-Reduce (GPSR) parsers [15,25] here, which "lift" the idea of textual Generalized LR (GLR) parsers 4 to HRGs and which can be used for any HRG. GPSR parsers manage not just one, but many stacks simultaneously, which are stored and processed in a so-called graph-structured stack (GSS) compressed in the form of a DAG.…”

Visualizing Visual Parser Execution

“…The proposed visualization and exploration approach can be used with many different types of visual parsers, either top-down or bottom-up. In fact, it has already been used with two different VGLR parser approaches based on [contextual] hyperedge replacement grammars ([C]HRGs) [13,9,24] and extended positional grammars [7,8].…”

“…The authors have realized several visual parsers and visual parser generator [7,8,9,13,24]. Developing VGLR parsers is challenging because one has to coordinate several non-trivial data structures like the input, the GSS, and the parse forest.…”