Labels:black | darkness OCR: Application Comparison Ratio Comments Software test and C test application: 272K lines, 22-47 C version implemented first. installation 120 engineer months. Tcl/Perl version replaced C FIS application: 90K lines, both C applications. 60 engineer months. Tcl/Perl version 7.7K lines, 8 engineer months Database library and C++ library: 2-3 months 8-12 C++ version implemented application Tel library: 1 week first; Tel version of library had C++ application: 2 months 60 more functionality. Tel application: 1 day Display oil well pro- C version: 3 months 6 Tel version implemented first. duction curves Tel version: 2 weeks Query dispatcher C version: 1200 lines, 4-8 2.5-8 C version implemented first, wecks uncommented. Tel version Tel version: 500 lines, 1 weck had comments, more func - tionality. Spreadsheet tool C version: 1460 lines 4 Tel version implemented first. Tel version: 380 Lines Simulator and GUI Java version: 3400 lines, 3-4 2-3 Tel version had 10-20% more wecks. functionality, was imple- Tcl version: 1600 lines, < 1 mented first. weck. Table 1. Each row of the table describes one or two applications that were implemented twice, once with a system programming language such as C or Java and once with a script- ing language such as Tel. The Ratio column gives the ratio of lines of code or develop- ment time for the two implementations (>1 means the system programming language required more effort). In most cases the two versions were implemented by different peo- ple. The information in the table was provided by various Tel developers in response to an article posted on the comp.lang.tel newsgroup. Scripting languages showed less improve- ment in situations where they were used for the first implementation, most likely because the second implementation benefited from the experiences of the first.
