[ https://issues.apache.org/jira/browse/HIVE-15743?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15840937#comment-15840937 ]
Sergey Shelukhin edited comment on HIVE-15743 at 1/27/17 4:21 AM: ------------------------------------------------------------------ cc [~mmccline] We can probably just c/p parts of FloatingDecimal - merge the parsing and doubleValue, and change them to operate on byte array. It only needs to recognize like 6-8 letters aside from normal numeric stuff (nan and infinity aside), so we should be safe since we always use utf8 was (Author: sershe): cc [~mmccline] We can probably just c/p parts of FloatingDecimal - merge the parsing and doubleValue, and change them to operate on byte array. It only needs to recognize like 6-8 letters aside from normal numeric stuff (nan and infinity aside), so we should be safe since we always use utf8 > vectorized text parsing: speed up double parse > ---------------------------------------------- > > Key: HIVE-15743 > URL: https://issues.apache.org/jira/browse/HIVE-15743 > Project: Hive > Issue Type: Bug > Reporter: Sergey Shelukhin > Attachments: tpch-without.png > > > {noformat} > Double.parseDouble( > new String(bytes, fieldStart, fieldLength, > StandardCharsets.UTF_8));{noformat} > This takes ~25% of the query time in some cases. -- This message was sent by Atlassian JIRA (v6.3.4#6332)