Original PR doesn't make clear why this is 900-now assuming we want to
cross the 14 bit length boundary.
Up to 900 creates a super large (>2^16) message. Not sure why that needs
testing.
Also remove the packs at lower sizes.
Signed-off-by: Miek Gieben <miek@miek.nl>
* Reduce compression memory use with map[string]uint16
map[string]uint16 uses 25% less memory per-entry than a map[string]int
(16+2)/(16+8) = 0.75. All entries in the compression map are bound by
maxCompressionOffset which is 14-bits and fits within a uint16.
* Add PackMsg benchmark with more RRs
* Add a comment to the compressionMap struct
* Pretty print test compression map differences
* Use compressionMapsDifference in TestPackDomainNameCompressionMap
This isn't strictly needed as it only contains a small number of
entries, but is consistent nonetheless.
* Fix map ordering in compressionMapsDifference
* Fix compressed length calculations for escaped names
* Add Len benchmark for escaped name
* Fix length with escaping after compression point
* Avoid calling escapedNameLen multiple times in domainNameLen
* Use regular quotes for question in TestMsgCompressLengthEscaped
* Remove fullSize return from compressionLenSearch
This wasn't used anywhere but TestCompressionLenSearch, and was very
wrong.
* Add generated compressedLen functions and use them
This replaces the confusing and complicated compressionLenSlice
function.
* Use compressedLenWithCompressionMap even for uncompressed
This leaves the len() functions unused and they'll soon be removed.
This also fixes the off-by-one error of compressedLen when a (Q)NAME
is ".".
* Use Len helper instead of RR.len private method
* Merge len and compressedLen functions
* Merge compressedLen helper into Msg.Len
* Remove compress bool from compressedLenWithCompressionMap
* Merge map insertion into compressionLenSearch
This eliminates the need to loop over the domain name twice when we're
compressing the name.
* Use compressedNameLen for NSEC.NextDomain
This was a mistake.
* Remove compress from RR.len
* Add test case for multiple questions length
* Add test case for MINFO and SOA compression
These are the only RRs with multiple compressible names within the same
RR, and they were previously broken.
* Rename compressedNameLen to domainNameLen
It also handles the length of uncompressed domain names.
* Use off directly instead of len(s[:off])
* Move initial maxCompressionOffset check out of compressionLenMapInsert
This should allow us to avoid the call overhead of
compressionLenMapInsert in certain limited cases and may result in a
slight performance increase.
compressionLenMapInsert still has a maxCompressionOffset check inside
the for loop.
* Rename compressedLenWithCompressionMap to msgLenWithCompressionMap
This better reflects that it also calculates the uncompressed length.
* Merge TestMsgCompressMINFO with TestMsgCompressSOA
They're both testing the same thing.
* Remove compressionLenMapInsert
compressionLenSearch does everything compressionLenMapInsert did anyway.
* Only call compressionLenSearch in one place in domainNameLen
* Split if statement in domainNameLen
The last two commits worsened the performance of domainNameLen
noticably, this change restores it's original performance.
name old time/op new time/op delta
MsgLength-12 550ns ±13% 510ns ±21% ~ (p=0.050 n=10+10)
MsgLengthNoCompression-12 26.9ns ± 2% 27.0ns ± 1% ~ (p=0.198 n=9+10)
MsgLengthPack-12 2.30µs ±12% 2.26µs ±16% ~ (p=0.739 n=10+10)
MsgLengthMassive-12 32.9µs ± 7% 32.0µs ±10% ~ (p=0.243 n=9+10)
MsgLengthOnlyQuestion-12 9.60ns ± 1% 9.20ns ± 1% -4.16% (p=0.000 n=9+9)
* Remove stray newline from TestMsgCompressionMultipleQuestions
* Remove stray newline in length_test.go
This was introduced when resolving merge conflicts.
* Use map[string]struct{} for compression map in Len
map[string]int requires 8 bytes per entry to store the unused position
information.
* Add MsgLength benchmark with more RRs
* Pass dns.Compress explicitly to packBufferWithCompressionMap
* Avoid creating compression map for question only Msg
This idea was inspired by:
"Skip dname compression for replies with no answers."
https://www.nlnetlabs.nl/bugs-script/show_bug.cgi?id=235
* Continue compressing multiple questions
* Fixed len computation when size just goes beyond 14 bits
* Added bouds checks around 14bits
* Len() always right including when around 14bits boudaries
* Avoid splitting into labels when not applicable
* Fixed comments
* Added comments in code
* Added new test cases
* Fixed computation of Len() for SRV and all kind of records
* Fixed Sign that was relying on non-copy for Unit tests
* Removed unused padding
* Fixed typo in PackBuffer() function
* Added comment about packBufferWithCompressionMap() for testing purposes
This offset max was not taking into account leading Len() to emit a
smaller value that could not be matched by Pack(), i.e all names can
be fully compressed or used as a target for compression.
Split length tests off in seperate file length_test.go to clean up
dns_test.og a bit.
Miek Gieben
Tom Thorogood
Pierre Souchay