Improve TypeError error message wording from config_set()

[fronders]

Describe the bug

UBXMessage.config_set() tries to lookup attribute types and fails because of val2bytes() function for att to valb conversion

Issue is caused by code on line 281:

pyubx2/src/pyubx2/ubxhelpers.py

Lines 265 to 297 in 1227e03

	def val2bytes(val, att: str) -> bytes:
	"""
	Convert value to bytes for given UBX attribute type.
	:param object val: attribute value e.g. 25
	:param str att: attribute type e.g. 'U004'
	:return: attribute value as bytes
	:rtype: bytes
	:raises: UBXTypeError
	"""
	if att == ubt.CH: # single variable-length string (e.g. INF-NOTICE)
	return val.encode("utf-8", "backslashreplace")
	atts = attsiz(att)
	if atttyp(att) in ("C", "X"): # byte or char
	valb = val
	elif atttyp(att) in ("E", "L", "U"): # unsigned integer
	valb = val.to_bytes(atts, byteorder="little", signed=False)
	elif atttyp(att) == "A": # array of unsigned integers
	atts = attsiz(att)
	valb = b""
	for i in range(atts):
	valb += val[i].to_bytes(1, byteorder="little", signed=False)
	elif atttyp(att) == "I": # signed integer
	valb = val.to_bytes(atts, byteorder="little", signed=True)
	elif att == ubt.R4: # single precision floating point
	valb = struct.pack("<f", val)
	elif att == ubt.R8: # double precision floating point
	valb = struct.pack("<d", val)
	else:
	raise ube.UBXTypeError(f"Unknown attribute type {att}")
	return valb

which returns integer instead of bytes in case of bitmask attribute.

Adding to_bytes() conversion fixes the issue.
This is especially critical for X8 type, where little-endian conversion is a must.
I don't see why X1/2/4/8 attributes need to be treated differently from U1/2/4/8 ones

Info

1. The pyubx2 version 1.2.48
2. The complete Python script.

import pyubx2

command = pyubx2.UBXMessage.config_set(
    layers=pyubx2.SET_LAYER_FLASH,
    transaction=pyubx2.TXN_NONE,
    cfgData=[(0x20920006,0)]
)
print(command)

3. The error message and full traceback.

Traceback (most recent call last):
  File "D:\test\ublox\ublox_test.py", line 3, in <module>
    command = pyubx2.UBXMessage.config_set(
  File "D:\test\.venv\lib\site-packages\pyubx2\ubxmessage.py", line 742, in config_set
    lis = lis + keyb + valb
TypeError: can't concat int to bytes

4. Expected Behaviour: code prints

<UBX(CFG-VALSET, version=0, ram=0, bbr=0, flash=1, action=0, reserved0=0, CFG_INFMSG_NMEA_I2C=b'\x00')>

To Reproduce

Steps to reproduce the behaviour:

1. Try to compose any CFG-VALSET item with "X1" or "X8" type

Desktop:

• Windows 11 x64.
• UART1 connection via USB-UART dongle.

GNSS/GPS Device:

• Device Model/Generation: ublox NEO-F10N
• Firmware Version: SPG 6.00
• Protocol: 40.00

[semuadmin]

Hi @fronders,

This is expected behaviour for X type attributes. The UBX protocol makes an intentional distinction between U (unsigned integer) and X (byte) types, and pyubx2 has to observe this distinction. Not all X type attributes can be interpreted as unsigned integers (the majority are actually character strings), so a .to_bytes() conversion would not be appropriate in all cases.

In this case the issue can be simply resolved by supplying the value as bytes, i.e.:

import pyubx2

command = pyubx2.UBXMessage.config_set(
    layers=pyubx2.SET_LAYER_FLASH,
    transaction=pyubx2.TXN_NONE,
    cfgData=[(0x20920006, b"\x00")],
)
print(command)

<UBX(CFG-VALSET, version=0, ram=0, bbr=0, flash=1, action=0, reserved0=0, CFG_INFMSG_NMEA_I2C=b'\x00')>

As to why some config database values are defined as X when a U (or C) type would seem more intuitive - I'm afraid you'll have to ask u-blox that question!

[fronders]

OK, fair enough, I see many X8 ones are char sequences, while X1, X2 and X4 mostly are bitmasks (which should be treated as unsigned imo). But at least the fact that you need to pass little-endian bytes to cfgData needs to be documented somewhere :) The val2bytes(val, att) behavior is a bit inconsistent then, as one would expect it to return bytes. Maybe check for incoming val type and raise an exception telling user to pass bytes instead of dying silently?

[semuadmin]

But at least the fact that you need to pass little-endian bytes to cfgData needs to be documented somewhere :)

The 'little-endian' nature of UBX payload bytes is documented in the u-blox Interface Specifications. What you need to pass to cfgData depends on the attribute type. The type associated with each individual configuration database key is documented in the relevant Interface Specification and can be referenced via ubxtypes_configdb.py.

The val2bytes(val, att) behavior is a bit inconsistent then, as one would expect it to return bytes.

The val2bytes() method does always return bytes as long as valid input arguments are provided.

Maybe check for incoming val type and raise an exception telling user to pass bytes instead of dying silently?

The pyubx2.config_set() function did not 'die silently' - it returned a TypeError which is the standard Python exception for - ahem - type errors. But I concede that the wording of the error message could be more helpful and I'm happy to update this to state explicitly which attribute 'type' it's expecting. I'll put it on the backlog and move this Issue to the Discussions Ideas and Feature Request forum.

[semuadmin]

Provisional updated TypeError wording:

...
TypeError: Attribute type X001 value 0 must be <class 'bytes'>, not <class 'int'>

[semuadmin]

Delivered in v1.2.49