[SOLR-12924] Optimize JavabinCodec to use byte[] backed CharSequence instead of String

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 8.0
• Assignee: Noble Paul
• Reporter: Noble Paul
• Created: 25/Oct/18 07:53
• Updated: 08/Jun/19 15:03
• Resolved: 25/Jan/19 03:27

Description

javabin format uses UTF8 format for String serialization and java Strings are UTF16 and java String always days a buffer copy of those UTF16 input.

So, what I propose is

• we create a byte[] backed CharSequence implementation say Utf8CharSequence.
• Optionally, javabinCodec can should be able to return this Utf8CharSequence instead of all Strings
• If a charAt() or toString() call is made to Utf8CharSequence it creates a String and cache it
• When this Utf8CharSequence needs to be serialized again in javabin format it should copy the underlying utf8 byte[]

Issue Links

• is related to: SOLR-12983 JavabinLoader should avoid creating String Objects and create UTF8CharSequence fields from byte[]
• is related to: SOLR-13130 during the ResponseBuilder.STAGE_GET_FIELDS avoid creation of Strings
• is related to: SOLR-13171 Make it possible to stream query result without creating java objects
• relates to: SOLR-12885 BinaryResponseWriter (javabin format) should directly copy from Bytesref to output