[ https://issues.apache.org/jira/browse/OPENNLP-1440?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17656472#comment-17656472 ]
ASF GitHub Bot commented on OPENNLP-1440: ----------------------------------------- mawiesne commented on code in PR #486: URL: https://github.com/apache/opennlp/pull/486#discussion_r1065418797 ########## opennlp-tools/src/main/java/opennlp/tools/cmdline/dictionary/DictionaryBuilderTool.java: ########## @@ -56,8 +58,8 @@ public void run(String[] args) { CmdLineUtil.checkInputFile("dictionary input file", dictInFile); CmdLineUtil.checkOutputFile("dictionary output file", dictOutFile); - try (InputStreamReader in = new InputStreamReader(new FileInputStream(dictInFile), encoding); - OutputStream out = new FileOutputStream(dictOutFile)) { + try (Reader in = new BufferedReader(new InputStreamReader(new FileInputStream(dictInFile), encoding)); + OutputStream out = new FileOutputStream(dictOutFile)) { Review Comment: This PR/issues targets _read_ IO operations, only. _Write_ is a separate topic. > Ensure files are read via buffered IO operations > ------------------------------------------------ > > Key: OPENNLP-1440 > URL: https://issues.apache.org/jira/browse/OPENNLP-1440 > Project: OpenNLP > Issue Type: Improvement > Components: Applications > Affects Versions: 2.1.0 > Reporter: Martin Wiesner > Assignee: Martin Wiesner > Priority: Minor > Fix For: 2.1.1 > > > Several classes in _opennnlp.tools_ exist which read files via > {{FileInputStream}} *without* using buffered IO. If IO is not buffered, this > can impose a (high) performance penalty as the JVM will have to use native > (JNI) calls more often (resulting in more sys-calls to the OS). > We can avoid that by adapting existing classes to use {{BufferedInputStream}} > or {{BufferedReader}} more consequently. -- This message was sent by Atlassian Jira (v8.20.10#820010)