This continues the process of converting the typeOfScript function to
use a combination of raw script analysis and the new tokenizer instead
of the far less efficient parsed opcodes.
In particular, it converts the detection of pay-to-pubkey-hash scripts
to use raw script analysis.
In order to accomplish this, it introduces two new functions. The first
one is named extractPubKeyHash and works with the raw script bytes
to simultaneously determine if the script is a pay-to-pubkey-hash script,
and in the case it is, extract and return the hash. The second new
function is named isPubKeyHashScript and is defined in terms of the
former.
The extract function approach was chosen because it is common for
callers to want to only extract relevant details from a script if the
script is of the specific type. Extracting those details requires
performing the exact same checks to ensure the script is of the correct
type, so it is more efficient to combine the two into one and define the
type determination in terms of the result so long as the extraction does
not require allocations.
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
----------------------------------------------------------------
BenchmarkIsPubKeyHashScript 165903 0.64 -100.00%
benchmark old allocs new allocs delta
----------------------------------------------------------------
BenchmarkIsPubKeyHashScript 1 0 -100.00%
benchmark old bytes new bytes delta
----------------------------------------------------------------
BenchmarkIsPubKeyHashScript 466945 0 -100.00%
This continues the process of converting the typeOfScript function to
use a combination of raw script analysis and the new tokenizer instead
of the far less efficient parsed opcodes.
In particular, it converts the detection of pay-to-alt-pubkey scripts to
use raw script analysis.
In order to accomplish this, it introduces two new functions. The first
one is named extractPubKeyAltDetails and works with the raw script bytes
to simultaneously determine if the script is a pay-to-alt-pubkey script,
and in the case it is, extract and return the relevant details. The
second new function is named isPubKeyAltScript and is defined in terms
of the former.
The extract function approach was chosen because it is common for
callers to want to only extract relevant details from a script if the
script is of the specific type. Extracting those details requires
performing the exact same checks to ensure the script is of the correct
type, so it is more efficient to combine the two into one and define the
type determination in terms of the result so long as the extraction does
not require allocations.
It is important to note that this new implementation intentionally
tightens the following semantics as compared to the existing
implementation:
- The signature type must now be one of the two supported types versus
allowing any single byte data push
- The public key must now be of the correct length for the given
signature type versus allowing any size up to 512 bytes
- The public key for schnorr secp256k1 pubkeys must now be a compressed
public key and adhere to the strict encoding requirements for them
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
---------------------------------------------------------------
BenchmarkIsAltPubKeyScript 143449 2.99 -100.00%
benchmark old allocs new allocs delta
---------------------------------------------------------------
BenchmarkIsAltPubKeyScript 1 0 -100.00%
benchmark old bytes new bytes delta
---------------------------------------------------------------
BenchmarkIsAltPubKeyScript 466944 0 -100.00%
This continues the process of converting the typeOfScript function to
use a combination of raw script analysis and the new tokenizer instead
of the far less efficient parsed opcodes.
In particular, it converts the detection of pay-to-pubkey scripts to use
raw script analysis.
In order to accomplish this, it introduces four new functions:
extractCompressedPubKey, extractUncompressedPubKey, extractPubKey, and
isPubKeyScript. The extractPubKey function makes use of
extractCompressedPubKey and extractUncompressedPubKey to combine their
functionality as a convenience and isPubKeyScript is defined in terms of
extractPubKey.
The extractCompressedPubKey works with the raw script bytes to
simultaneously determine if the script is a pay-to-compressed-pubkey
script, and in the case it is, extract and return the raw compressed
pubkey bytes.
Similarly, the extractUncompressedPubKey works in the same way except it
determines if the script is a pay-to-uncompressed-pubkey script and
returns the raw uncompressed pubkey bytes in the case it is.
The extract function approach was chosen because it is common for
callers to want to only extract relevant details from a script if the
script is of the specific type. Extracting those details requires
performing the exact same checks to ensure the script is of the correct
type, so it is more efficient to combine the two into one and define the
type determination in terms of the result so long as the extraction does
not require allocations.
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
------------------------------------------------------------
BenchmarkIsPubKeyScript 124749 4.01 -100.00%
benchmark old allocs new allocs delta
------------------------------------------------------------
BenchmarkIsPubKeyScript 1 0 -100.00%
benchmark old bytes new bytes delta
------------------------------------------------------------
BenchmarkIsPubKeyScript 466944 0 -100.00%
This continues the process of converting the typeOfScript function to
use a combination of raw script analysis and the new tokenizer instead
of the far less efficient parsed opcodes.
In particular, for this commit, since the ability to detect multisig
scripts via the new tokenizer is now available, the function is simply
updated to make use of it.
This begins the process of converting the typeOfScript function to use a
combination of raw script analysis and the new tokenizer instead of the
far less efficient parsed opcodes with the intent of significantly
optimizing the function.
In order to ease the review process, each script type will be converted
in a separate commit and the typeOfScript function will be updated such
that the script is only parsed as a fallback for the cases that are not
already converted to more efficient raw script variants.
In particular, for this commit, since the ability to detect
pay-to-script-hash via raw script analysis is now available, the
function is simply updated to make use of it.
This converts the typeOfScript function to accept a script version and
raw script instead of an array of internal parsed opcodes in order to
make it more flexible for raw script analysis.
Also, this adds a comment to CalcScriptInfo to call out the specific
version semantics and deprecates the function since nothing currently
uses it, and the relevant information can now be obtained by callers
more directly through the use of the new script tokenizer.
All other callers are updated accordingly.
This converts the GetPreciseSigOpCount function to use a combination of
raw script analysis and the new tokenizer instead of the far less
efficient parseScript thereby significantly optimizing the function.
In particular it uses the recently converted isScriptHashScript,
IsPushOnlyScript, and countSigOpsV0 functions along with the recently
added finalOpcodeData functions.
It also modifies the comment to explicitly call out the script version
semantics.
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
------------------------------------------------------------------
BenchmarkGetPreciseSigOpCount 287939 1077 -99.63%
benchmark old allocs new allocs delta
------------------------------------------------------------------
BenchmarkGetPreciseSigOpCount 3 0 -100.00%
benchmark old bytes new bytes delta
------------------------------------------------------------------
BenchmarkGetPreciseSigOpCount 934657 0 -100.00%
This moves the check for non push-only pay-to-script-hash signature
scripts before the script parsing logic when creating a new engine
instance to avoid the extra overhead in the error case.
This modifies the check for whether or not a pay-to-script-hash
signature script is a push only script to make use of the new and more
efficient raw script function.
Also, since the script will have already been checked further above when
the ScriptVerifySigPushOnly flags is set, avoid checking it again in
that case.
This converts the IsPushOnlyScript function to make use of the new
tokenizer instead of the far less efficient parseScript thereby
significantly optimizing the function.
It also deprecates the isPushOnly function that requires opcodes in
favor of the new function and modifies the comment on IsPushOnlyScript
to explicitly call out the script version semantics.
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
---------------------------------------------------------------
BenchmarkIsPayToScriptHash 139961 0.66 -100.00%
benchmark old allocs new allocs delta
---------------------------------------------------------------
BenchmarkIsPayToScriptHash 1 0 -100.00%
benchmark old bytes new bytes delta
---------------------------------------------------------------
BenchmarkIsPayToScriptHash 466944 0 -100.00%
This converts the isAnyKindOfScriptHash function to analyze the raw
script instead of requiring far less efficient parsed opcodes thereby
significantly optimizing the function.
Since the function relies on isStakeScriptHash to identify a stake
tagged pay-to-script-hash, and is the only consumer of it, this also
converts that function to analyze the raw script and renames it to
isStakeScriptHashScript for more consistent naming.
Finally, the tests are updated accordingly.
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
-------------------------------------------------------------------
BenchmarkIsAnyKindOfScriptHash 101249 3.83 -100.00%
benchmark old allocs new allocs delta
-------------------------------------------------------------------
BenchmarkIsAnyKindOfScriptHash 1 0 -100.00%
benchmark old bytes new bytes delta
-------------------------------------------------------------------
BenchmarkIsAnyKindOfScriptHash 466944 0 -100.00%
This adds tests to ensure the isAnyKindOfScriptHash function properly
identifies the four stake-tagged pay-to-script-hash possibilities in
addition to ensuring they are not misidentified as standard
pay-to-script-hash scripts.
This converts the GetSigOpCount function to make use of the new
tokenizer instead of the far less efficient parseScript thereby
significantly optimizing the function.
A new function named countSigOpsV0 which accepts the raw script is
introduced to perform the bulk of the work so it can be reused for
precise signature operation counting as well in a later commit. It
retains the same semantics in terms of counting the number of signature
operations either up to the first parse error or the end of the script
in the case it parses successfully as required by consensus.
Finally, this also deprecates the getSigOpCount function that requires
opcodes in favor of the new function and modifies the comment on
GetSigOpCount to explicitly call out the script version semantics.
The following is a before and after comparison of analyzing a large
script:
benchmark old ns/op new ns/op delta
-----------------------------------------------------------
BenchmarkGetSigOpCount 163896 1048 -99.36%
benchmark old allocs new allocs delta
-----------------------------------------------------------
BenchmarkGetSigOpCount 1 0 -100.00%
benchmark old bytes new bytes delta
-----------------------------------------------------------
BenchmarkGetSigOpCount 466945 0 -100.00%
This converts the IsMultisigSigScript function to analyze the raw script
and make use of the new tokenizer instead of the far less efficient
parseScript thereby significantly optimizing the function.
In order to accomplish this, it first rejects scripts that can't
possibly fit the bill due to the final byte of what would be the redeem
script not being the appropriate opcode or the overall script not having
enough bytes. Then, it uses a new function that is introduced named
finalOpcodeData that uses the tokenizer to return any data associated
with the final opcode in the signature script (which will be nil for
non-push opcodes or if the script fails to parse) and analyzes it as if
it were a redeem script when it is non nil.
It is also worth noting that this new implementation intentionally has
the same semantic difference from the existing implementation as the
updated IsMultisigScript function in regards to allowing zero pubkeys
whereas previously it incorrectly required at least one pubkey.
Finally, the comment is modified to explicitly call out the script
version semantics.
The following is a before and after comparison of analyzing a large
script that is not a multisig script and both a 1-of-2 multisig public
key script (which should be false) and a signature script comprised of a
pay-to-script-hash 1-of-2 multisig redeem script (which should be true):
benchmark old ns/op new ns/op delta
-----------------------------------------------------------------------
BenchmarkIsMultisigSigScriptLarge 158149 4 -100.00%
BenchmarkIsMultisigSigScript 3445 202 -94.14%
benchmark old allocs new allocs delta
-----------------------------------------------------------------------
BenchmarkIsMultisigSigScriptLarge 9 0 -100.00%
BenchmarkIsMultisigSigScript 3 0 -100.00%
benchmark old bytes new bytes delta
-----------------------------------------------------------------------
BenchmarkIsMultisigSigScriptLarge 533189 0 -100.00%
BenchmarkIsMultisigSigScript 9472 0 -100.00%
This converts the IsMultisigScript function to make use of the new
tokenizer instead of the far less efficient parseScript thereby
significantly optimizing the function.
In order to accomplish this, it introduces two new functions. The first
one is named extractMultisigScriptDetails and works with the raw script
bytes to simultaneously determine if the script is a multisignature
script, and in the case it is, extract and return the relevant details.
The second new function is named isMultisigScript and is defined in
terms of the former.
The extract function accepts the script version, raw script bytes, and a
flag to determine whether or not the public keys should also be
extracted. The flag is provided because extracting pubkeys results in
an allocation that the caller might wish to avoid.
The extract function approach was chosen because it is common for
callers to want to only extract relevant details from a script if the
script is of the specific type. Extracting those details requires
performing the exact same checks to ensure the script is of the correct
type, so it is more efficient to combine the two into one and define the
type determination in terms of the result so long as the extraction does
not require allocations.
It is important to note that this new implementation intentionally has a
semantic difference from the existing implementation in that it will now
correctly identify a multisig script with zero pubkeys whereas
previously it incorrectly required at least one pubkey. This change is
acceptable because the function only deals with standardness rather than
consensus rules.
Finally, this also deprecates the isMultiSig function that requires
opcodes in favor of the new functions and deprecates the error return on
the export IsMultisigScript function since it really does not make sense
given the purpose of the function.
The following is a before and after comparison of analyzing both a large
script that is not a multisig script and a 1-of-2 multisig public key
script:
benchmark old ns/op new ns/op delta
-------------------------------------------------------------------
BenchmarkIsMultisigScriptLarge 121599 8.63 -99.99%
BenchmarkIsMultisigScript 797 72.8 -90.87%
benchmark old allocs new allocs delta
-------------------------------------------------------------------
BenchmarkIsMultisigScriptLarge 1 0 -100.00%
BenchmarkIsMultisigScript 1 0 -100.00%
benchmark old bytes new bytes delta
-------------------------------------------------------------------
BenchmarkIsMultisigScriptLarge 466944 0 -100.00%
BenchmarkIsMultisigScript 2304 0 -100.00%
This converts the IsPayToScriptHash function to analyze the raw script
instead of using the far less efficient parseScript thereby
significantly optimizing the function.
In order to accomplish this, it introduces two new functions. The first
one is named extractScriptHash and works with the raw script bytes to
simultaneously determine if the script is a p2sh script, and in the case
it is, extract and return the hash. The second new function is named
isScriptHashScript and is defined in terms of the former.
The extract function approach was chosen because it is common for
callers to want to only extract relevant details from a script if the
script is of the specific type. Extracting those details requires
performing the exact same checks to ensure the script is of the correct
type, so it is more efficient to combine the two into one and define the
type determination in terms of the result so long as the extraction does
not require allocations.
Finally, this also deprecates the isScriptHash function that requires
opcodes in favor of the new functions and modifies the comment on
IsPayToScriptHash to explicitly call out the script version semantics.
The following is a before and after comparison of analyzing a large
script that is not a p2sh script:
benchmark old ns/op new ns/op delta
---------------------------------------------------------------
BenchmarkIsPayToScriptHash 139961 0.66 -100.00%
benchmark old allocs new allocs delta
---------------------------------------------------------------
BenchmarkIsPayToScriptHash 1 0 -100.00%
benchmark old bytes new bytes delta
---------------------------------------------------------------
BenchmarkIsPayToScriptHash 466944 0 -100.00%
This converts the isStakeOpcode function to accept an opcode as a byte
instead of the internal opcode data struct in order to make it more
flexible for raw script analysis.
It also updates all callers accordingly.
This converts the asSmallInt function to accept an opcode as a byte
instead of the internal opcode data struct in order to make it more
flexible for raw script analysis.
It also updates all callers accordingly.
This converts the isSmallInt function to accept an opcode as a byte
instead of the internal opcode data struct in order to make it more
flexible for raw script analysis.
The comment is modified to explicitly call out the script version
semantics.
Finally, it updates all callers accordingly.
This converts the tests for calculating signature hashes to use the
exported function which handles the raw script versus the now deprecated
variant requiring parsed opcodes.
This modifies the CalcSignatureHash function to make use of the new
signature hash calculation function that accepts raw scripts without
needing to first parse them. Consequently, it also doubles as a slight
optimization to the execution time and a significant reduction in the
number of allocations.
In order to convert the CalcScriptHash function and keep the same
semantics, a new function named checkScriptParses is introduced which
will quickly determine if a script can be fully parsed without failure
and return the parse failure in the case it can't.
The following is a before and after comparison of analyzing a large
multiple input transaction:
benchmark old ns/op new ns/op delta
-------------------------------------------------------
BenchmarkCalcSigHash 2792057 2760042 -1.15%
benchmark old allocs new allocs delta
-------------------------------------------------------
BenchmarkCalcSigHash 1691 1068 -36.84%
benchmark old bytes new bytes delta
-------------------------------------------------------
BenchmarkCalcSigHash 521673 438604 -15.92%
This introduces a new function named calcSignatureHashRaw which accepts
the raw script bytes to calculate the script hash versus requiring the
parsed opcode only to unparse them later in order to make it more
flexible for working with raw scripts.
Since there are several places in the rest of the code that currently
only have access to the parsed opcodes, this modifies the existing
calcSignatureHash to first unparse the script before calling the new
function.
Note that the code in the signature hash calculation to remove all
instances of OP_CODESEPARATOR from the script is removed because that is
a holdover from BTC code which does not apply to v0 Decred scripts since
OP_CODESEPARATOR is completely disabled in Decred and thus there can
never actually be one in the script.
Finally, it removes the removeOpcode function and related tests since it
is no longer used.
This converts the DisasmString function to make use of the new
zero-allocation script tokenizer instead of the far less efficient
parseScript thereby significantly optimizing the function.
In order to facilitate this, the opcode disassembly functionality is
split into a separate function called disasmOpcode that accepts the
opcode struct and data independently as opposed to requiring a parsed
opcode. The new function also accepts a pointer to a string builder so
the disassembly can be more efficiently be built.
While here, the comment is modified to explicitly call out the script
version semantics.
The following is a before and after comparison of a large script:
benchmark old ns/op new ns/op delta
----------------------------------------------------------
BenchmarkDisasmString 288729 94157 -67.39%
benchmark old bytes new bytes delta
----------------------------------------------------------
BenchmarkDisasmString 584611 177528 -69.63%
This implements an efficient and zero-allocation script tokenizer that
is exported to both provide a new capability to tokenize scripts to
external consumers of the API as well as to serve as a base for
refactoring the existing highly inefficient internal code.
It is important to note that this tokenizer is intended to be used in
consensus critical code in the future, so it must exactly follow the
existing semantics.
The current script parsing mechanism used throughout the txscript module
is to fully tokenize the scripts into an array of internal parsed
opcodes which are then examined and passed around in order to implement
virtually everything related to scripts.
While that approach does simplify the analysis of certain scripts and
thus provide some nice properties in that regard, it is both extremely
inefficient in many cases, and makes it impossible for external
consumers of the API to implement any form of custom script analysis
without manually implementing a bunch of error prone tokenizing code or,
alternatively, the script engine exposing internal structures.
For example, as shown by profiling the total memory allocations of an
initial sync, the existing script parsing code allocates a total of
around 295.12GB, which equates to around 50% of all allocations
performed. The zero-alloc tokenizer this introduces will allow that to
be reduced to virtually zero.
The following is a before and after comparison of tokenizing a large
script with a high opcode count using the existing code versus the
tokenizer this introduces for both speed and memory allocations:
benchmark old ns/op new ns/op delta
------------------------------------------------------------
BenchmarkScriptParsing 153099 961 -99.37%
benchmark old allocs new allocs delta
------------------------------------------------------------
BenchmarkScriptParsing 1 0 -100.00%
benchmark old bytes new bytes delta
------------------------------------------------------------
BenchmarkScriptParsing 466945 0 -100.00%
The following is an overview of the changes:
- Introduce new error code ErrUnsupportedScriptVersion
- Implement zero-allocation script tokenizer
- Add a full suite of tests to ensure the tokenizer works as intended
and follows the required consensus semantics
- Add an example of using the new tokenizer to count the number of
opcodes in a script
- Update README.md to include the new example
- Update script parsing benchmark to use the new tokenizer
This deprecates the GetMultisigMandN function which should never have
been added since the CalcMultiSigStats function already existed for this
purpose.
While here, redefine the function in terms of CalcMultiSigStats.
This function is only useful for internal consensus purposes within the
script engine and as such should not be exported.
While here, also add a comment to specify to the script version
semantics.
This adds the go 1.11 directive to all of the modules in order to
clearly mark they build and work with that version. Go 1.12 modified
the tools such that tidy will automatically add the new version to
modules that do not already have a directive and that would prevent
builds on Go 1.11 through Go 1.11.3 which is not desirable.
As is already well commented in the code, the sequence number parameter
of the CHECKSEQUENCEVERIFY opcode requires 5 bytes instead of the
standard 4 bytes allowed by math opcodes. This introduces a constant
for the value instead of hardcoding 5 to increase readability and
potentially allow the value to be exported in the future.
