| [henshin-dev] Support for EMap in Matching Engine |
Dear all, I yesterday committed a patch to the Henshin Interpreter Engine to exploit EMap references to speed up the matching. In EMF, you can create special references which use an EMap as data structure and then use a key to quickly match the value without iterating over all objects. See the EMF wiki how to use this in your meta model. (http://wiki.eclipse.org/index.php/EMF-FAQ#How_do_I_create_a_Map_in_EMF.3F) I also created an example (as part of the examples-plugin) which is shown in the attachment. It is based on the original Bank example. But this version uses the EMap-references for a quick matching of Accounts based on the accountIDs as keys. With the new support for EMap references in the interpreter, executing the transferMoney-rule 10.000 times on a bank with 100.000 accounts is 3 seconds compared with 300 seconds without the support. However, it is important that the ordering of the variables for the interpreter is in such a way that the EMap-node is matched before the „value“ nodes (in the banking example the Bank2AccountMap-node before the Account-node). This is currently not well supported by the automatic variable optimization, so you might switch it off and order the nodes manually in the tree editor in a good way which i also had to do for the new banking example. Cheers, Matthias
Attachment:
bankmap_henshin.pdf
Description: bankmap_henshin.pdf
Attachment:
bankmap.pdf
Description: bankmap.pdf