Package edu.illinois.ncsa.isda.softwareserver.polyglot.weighted

Polyglot tools supporting weighted and parameterized I/O-graphs.

See: Description

Package edu.illinois.ncsa.isda.softwareserver.polyglot.weighted Description

Polyglot tools supporting weighted and parameterized I/O-graphs. All the magic is in IOGraphWeightsToolHeadless and the configuration file IOGraphWeightsToolHeadless.conf. IOGraphWeightsToolHeadless contains a main that will start the whole process. You start the program with multiple tuples of arguments, a folder and an extension, for example ~/Pictures jpg ~/Pictures png (4 arguments in total in this case). It will run through the following steps: 1) create an iograph based on the software servers and CSR specified in the configuration file. 2) take each tuple and check the files, it will recurse into the folder specified and find all files with the specific extension in the folder or subfolders and add them to the list. Next it will compute the number of direct conversions you can do from A->B and B->A, for example jpg->png and png->jpg. (see note 1) 3) put all jobs in the queue, queue is not being processed yet. Each job will be all files converted from A->B->A`, using a specific set of software for conversions (see note 1). 4) create folders to hold test outputs, folder 0 will hold the original files, all other conversions will have their own folder which will hold the results of a conversion from A->B->A`. 5) copy the original files to folder 0 and add all conversions to an executer thread, the executor thread will run as many jobs in parallel as there are software servers. 6) each conversion thread will run 1 conversion at a time, saving the result in a special folder. 7) once all conversions in this thread are done it will add all results to the compare executor, each result will be compared based on the comparisons specified in the config file. The compare executor will have a single thread. 8) the comparison thread will upload the original file and the resulting file to versus (if not uploaded yet) and use the measure info to compare the two datasets. The results are uploaded back to ICR (in uploadResult). This requires a modified version of ICR, but we might be able to replace this with a function that just writes the results to disk. The measure result will be made positive with a ceiling of 9999. 10000 == bad measure, 20000 == no result found, 30000 == conversion failed. 9) once all conversions and comparisons are done the system will exit. Note 1: if there are 2 pieces of software (S1, S2) that do the conversion from A->B and 2 (S3, S4) that can do the conversion from B->A there will be a total of 4 different permutations tested.

Copyright © 2016. All rights reserved.