v5.4.0

v5.4.0 (2025-10-27)

Added

• Add validation for ABI method names to match ARC-4 specification (^[_A-Za-z][A-Za-z0-9_]*$).

• Added support for configuration of language server log level and debounce interval.

v5.3.2

v5.3.2 (2025-10-20)

Fixed

• Fixed incorrect validation for dynamic structs ending with a bool.

v5.3.1

v5.3.1 (2025-10-20)

Fix

• correct typo in --validate-abi-return CLI option (a194faa)

v5.3.0

v5.3.0 (2025-10-20)

Added

• Added validate_encoding option to algopy.arc4.abimethod, this allows controlling whether ABI arguments for the method are validated or not
• Added .validate() method to ARC4 and native data types to validate a value is encoded correctly
• Added support for validating dynamic types
• Added --validate-abi-args and --validate-abi-return options for more granular control of automatic ABI validation
• Added output to display puyapy version when compiling

Changed

• Preserve asserts with error messages followed by unconditional approval.

Deprecated

• The --validate-abi-dynamic-severity option has been deprecated since dynamic types can now also be validated
• The --validate-abi-values option has been deprecated, use --validate-abi-args and --validate-abi-return instead

v4.11.0

v4.11.0 (2025-10-20)

Added

• Added validate_encoding option to algopy.arc4.abimethod, this allows controlling whether ABI arguments for the method are validated or not
• Added .validate() method to ARC4 and native data types to validate a value is encoded correctly
• Added support for validating dynamic types
• Added --validate-abi-args and --validate-abi-return options for more granular control of automatic ABI validation
• Added output to display puyapy version when compiling

Fixed

• Prevent a critical error when optimizing extract ops with constant offsets and sizes

v5.2.0

v5.2.0 (2025-10-13)

Added

• Added a compiler option to validate the sizes of ARC-4 arguments to ensure they are well-formed this option is enabled by default
• Added a compiler option to control the severity of compiler output when notifying that nested dynamic types are not validated

v5.1.0

v5.1.0 (2025-10-13)

Added

• added language server code fix for .copy() requirements

v5.0.1

v5.0.1 (2025-10-06)

Fixed

• creating a type alias from another type alias is now supported e.g. Timestamp = algopy.arc4.UInt64

• improved language server support for PyCharm LSP4IJ plugin

v5.0.0

v5.0.0 (2025-10-03)

PuyaPy 5.0 contains breaking changes, please refer to the Changed section and migration guide for more details.

Added

• Documentation for how to get started with contributing a new front end langauge for the compiler.

• Added algopy.FixedArray that supports mutable, native and dynamic elements in an array with a fixed size.

• Added algopy.Struct that supports mutable, native and dynamic members.

• Added .to_native(element_type) function for converting algopy.arc4.DynamicArray to algopy.Array and algopy.arc4.StaticArray to algopy.FixedArray.

• Added a PuyaPy CLI option and @abimethod argument resource_encoding that controls how
  resource types (algopy.Asset, algopy.Application and algopy.Account) are handled in the
  ABI router.
  When set to value (the default) the types will use their underlying value type (uint64 or address).
  When set to index the types will use their ARC-4 reference aliases and when calling
  require populating the appropriate foreign array.
  This affects TEAL, AVM bytecode, ARC-32 and ARC-56 file output.

• Boxes containing fixed-size types larger than 4K can now be created/read/updated, provided
  the size of the innermost element being read/written is not larger than 4K.

• Improved reads and writes to nested fixed-size data types by reducing number of bounds checks
  where possible to do so safely.

• Added algopy.zero_bytes() function to easily initialise fixed-size bytes-backed types to
  all zero bytes.

• The optimiser can now detect when a storage key (box/global/local) is guaranteed to exist and
  can eliminate some redundant asserts.

• Repeated operations of concatenation, addition, multiplication, and uint64 bitwise operations
  involving two or more constants can now be optimized.

• Improved detection and elimination of redundant ARC-4 encode/decode operations, particularly
  those involving nested data structures, or where data structures have equivalent encoding but
  not identical types.

• Improved ARC-4 encoding generation for tuples with dynamically-sized elements.

• Added constant optimizations for sha3_256, sha512_256 and keccak256 op codes.

• Added support for reading and writing at an index or iterating a dynamic array of fixed size elements
  in a box > 4k in size.

• Added an optimization for converting substring3 with a len as the length argument to an extract START 0.

• Added an optimisation to eliminate extract_uint64 when it's consumed only from an itob.

• Optimize op arg to arg_N when N <= 3.

• Include an error message in TEAL, ARC-56 and debug outputs for when a dynamic array of fixed elements when writing to an out-of-bounds index.

• Added optimisation to always inline a subroutine that is a single intrinsic.

• Minor improvements to TEAL peephole optimisations.

• Support matching against multiple values in Python match-cases with |.

• Add .as_uint64() and .as_biguint() methods for all arc4.UIntN and arc4.BigUIntN classes to convert them to the primitive values. .as_uint64() works even for types with bytes size > 64 such as BigUIntN[typing.Literal[128]], BigUIntN[typing.Literal[512]], .etc, as long as the actual value fits within the bound of uint64. .as_uint64() throws overflow error if the value is larger than uint64

• Added extract, replace, resize, and splice methods to Box class

• Added AVM 13 support

• Added support for ops from langspec v4.3.0, including txn RejectVersion and app_params_get AppVersion

Changed

• algopy.Array can now be used in storage and allows native, mutable and dynamically sized types,
  however it no longer has reference semantics and may require .copy() when aliasing the array or it's contents.

  To retain the original behaviour of algopy.Array use the new algopy.ReferenceArray type.

• Require .copy()when extending or concatenating a sequence containing mutable values to maintain idiomatic compatibility.

• The default behaviour for the new resource_encoding option is different to prior versions, resource
  types (algopy.Asset, algopy.Application and algopy.Account) will now appear as their underlying
  ARC-4 value types in ABI methods (uint64 and address).
  This affects TEAL, AVM bytecode, ARC-32 and ARC-56 file output.

• Updated vendored mypy version from v1.13.0 to v1.17.0.

• Increase severity from warning to error when referencing logic sig only args in a contract and vice versa.

• Disable some lower level optimisations when running at O0.

• Improved performance of puya backend.

• ARC-4 router code generation has been changed to be more program-size efficient.

• Multiple other improvements to decrease program size have been introduced.

• Improved the help output for puyapy.

• Removed automatic output of ARC-32 app spec files.
  ARC-32 has been marked for eventual deprecation, with ARC-56 as the replacement.
  ARC-56 app spec files are already automatically output and continue to be.
  To continue outputting an ARC-32 file, specify the command line option --output-arc32.

• Changed constructors of ImmutableArray and ReferenceArray to take an Iterable instead of variadic args for consistency with other native array types.

• The default AVM version has been updated from 10 to 11 to match the current mainnet version.
  AVM version 10 remains a supported option for now.

Deprecated

• .native property in arc4.UIntN and arc4.BigUIntN classes has been depreacted. Use .as_uint64() or .as_biguint() instead.

• .ref property in Box class has been deprecated. The methods previously accessed via .ref e.g. extract, replace, resize, and splice are now available directly in the Box class.

• BoxRef class has been deprecated. Use Box[Bytes] instead.

Fixed

• Prevent a critical error when optimizing extract ops with constant offsets and sizes.

• Fixed a potential semantic compatibility bug when instantiating a new algopy.arc4.Struct
  using keyword arguments. The arguments were being evaluated in the order of the fields on the
  Struct class, rather than from left to right when being passed.

• Fixed a compilation failure when using --locals-coalescing-strategy=aggressive option in
  combination with algopy.ReferenceArray.

• Fixed a compilation failure when subroutines contained extremely long control flow sequences.

• Fixed a compilation failure when attempting to iterate an empty tuple.

• Prevent compilation failure in certain scenarios when performing copy-propagation.

• Fixed an issue where comparisons
  between boolean evaluations of Asset or Application types would produce unexpected results.

• No longer allow optimizations for setbit where the value could result in something other than a 0 or 1.

• Fixed typing for methods that should have only allowed bool inputs, specifically:

  • algopy.op.setbit
  • algopy.op.getbit
  • algopy.op.ITxnCreate.set_config_asset_default_frozen
  • algopy.op.ITxnCreate.set_freeze_asset_frozen
  • algopy.op.ITxnCreate.set_nonparticipation
  • algopy.itxn.InnerTransaction's nonparticipation arg
  • algopy.itxn.KeyRegistration's nonparticipation arg

• algopy.itxn.KeyRegistration no longer has any required fields since some combinations are mutually exclusive.

• Fix a compilation failure occurring at optimisation-level zero when a function that is
  declared to produce a result but always terminates the program is explicitly inlined.

• Fix an issue where a subroutine that never returns could be generated with ops that will
  always fail at runtime (specifically, using frame_dig and/or frame_bury in a subroutine
  that is detected as not requiring a proto instruction).

• Tuple types are now supported in template variables.

• Fixed an issue whereby reading the target variable of a urange based for-loop outside of the loop would not match the expected behaviour of the native Python equivalent range.

• Fixed an issue of semantic incompatibility concerning the order of evaluation of arguments when constructing named tuples.
  Previously, arguments were evaluated in the order of field declarations in the named tuple type, rather than in the order they appear in the instantiating call. This could cause different results when arguments have side effects.

v5.0.0-rc.14

v5.0.0-rc.14 (2025-09-26)

Fix

• handle relative imports correctly for language server dependency analysis (6788686)

Documentation

• fix typos and grammatical issues in docs (#564) (13cce9a)