[forge-issues] [JBoss JIRA] (FORGE-699) create a plugin for WildcardImportResolver

create a plugin for WildcardImportResolver ------------------------------------------ Key: FORGE-699 URL: https://issues.jboss.org/browse/FORGE-699 Project: Forge Issue Type: Story Components: Parsers / File Manipulation Affects Versions: 1.1.1.Final Reporter: fiorenzo pizza Assignee: fiorenzo pizza Fix For: 1.1.2.Final general notes We need to create some WildcardImportResolver (Responsible for providing additional import resolution functionality for situations where classes have referenced a package or wild-card import) summary: - when start creation of index --> new-project - when update the index --> re-open project | add|remove plugins to the project | add|remove maven dependencies | add|remove new class inside forge project - where save the index --> always - how information to save for future search --> className, jarName (null if the class is in the project), packageName, absoluteClassName - how search inside the index --> (ram hashmap? ram db?) - how generate new entry for the index --> cdi event general considerations 1) when forge create a new-project we should create some structure to add all classes used referenced by project 2) we need to add all classes required from some plugin (persistence|scaffolding..) 2) we need to add all classes inside maven dependencies 3) we need to add all classes created in the project 4) we need remove all deleted classes 5) we need to remove all classes inside dependecies removed 6) we need to update the index files, when forge project is reopen (the user could modify the project out of forge) problems 1) with a lot of jars, the indexing process should be slow!! 2) how synchronize the access to index 3) how we should synchronize of index update? requirements 1) we need some fast library to extract classes from jar files 2) we should use cdi event to start a single jar indexing HOW CREATE INDEX FROM JAR: possibile libraries: jandex, tattletale, jarindexer FIRST CANDIDATE: JANDEX i started to hack jandex code, because jandex is very fast. Jandex opens the jar files using java jar utilities: JarFile jar = new JarFile(jarFile); and for every class generate a ClassInfo object: ClassInfo info = indexer.index(jar.getInputStream(entry)); For my test, i removed some methods inside jandex: we don't care about annotations inside classes. Also i removed the generation of file index. questions: 1) why don't use H2, the Java SQL database, for memorize all informations about classes? How memorize for a single class: 1) classname, jar name (with version), package name, complete name