You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@commons.apache.org by "Henri Biestro (Jira)" <ji...@apache.org> on 2021/06/07 13:15:06 UTC

[jira] [Closed] (JEXL-277) Speedup arithmetic evaluations

     [ https://issues.apache.org/jira/browse/JEXL-277?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Henri Biestro closed JEXL-277.
------------------------------

> Speedup arithmetic evaluations
> ------------------------------
>
>                 Key: JEXL-277
>                 URL: https://issues.apache.org/jira/browse/JEXL-277
>             Project: Commons JEXL
>          Issue Type: Improvement
>    Affects Versions: 3.1
>         Environment: The [PR 15|https://github.com/apache/commons-jexl/pull/15] is prepared for this ticket
>            Reporter: Dmitri Blinov
>            Assignee: Henri Biestro
>            Priority: Minor
>             Fix For: 3.2
>
>
> As of now, Jexl inefficiently evaluates operators *+*, *-*, */*, *%*, *\** on Integer and Long types as it casts operands to BigInteger type, applies operation and casts the result back. This for example, creates 3 BigInteger instances for a simple operation of 1+1. All evaluations on Integer types, and in some cases, on Long types, can be performed on corresponding primitive *long* values, which is a magnitute of 200%-300% faster, in my measurments, and unneeded objects are not allocated on heap. Another point is to make selfAssign operators *overridable* in descendands of JexlArithmetic, instead of being able to be *oveloaded*, which can save the time by avoiding reflection when using syntax like {code}x += 1{code}.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)