CODEWITHSUNDEEP
tcl
my_question
my_question

Reputation: 3235

How to create a TCL variable of type bytearray

I am using TCL 8.4.20.

So I have the following code:

set a [binary format H2 1]
set b [binary format H2 2]
set c [binary format H2 3]
set bytes $a
append bytes $a
append bytes $b
append bytes $c

puts $bytes

I set a breakpoint at Tcl_PutsObjCmd() function in TCL's C source code and I see its argument, $bytes, is of type string while I expect it to be bytearray.

Question 1:Why is that? From the first assignment to the final appending, "bytes" accepts nothing but binary data.

The reason I do this experiment is, we have a TCL extension command in C, it expects the command argument is of byte array type - it has a check the value's typePtr should be tclByteArrayType. My TCL code currently fails on this command because the data passed to the command is of type string, just as demo'ed above.

I googled around, seems the "right" way to make a byte array object is to have every byte ready first and finally use one "binary format" command to put all into one. But it is a fairly big change to my current TCL code.

Question 2: Given that I already have a TCL variable whose data are all binaries (created using "binary format" for each byte and put together using "append") while its type is string, How can I change its internal type to "bytearray" through some TCL maneuvering?

Upvotes: 0

Views: 977

Answers (1)

schlenk
schlenk

Reputation: 7247

Technically, the internal type is not a guaranteed property. Everything is a string. The code may shimmer a type away whenever it feels like. And code that depends on the internal type is usually very brittle or outright broken.

So your C code should call Tcl_GetByteArrayFromObj() instead of peeking at the arguments internals. That does the proper conversion if the object has not yet a byteArray representation.

About your questions:

Why doesn't append of two byte arrays keep the byte array type?

It does, at least for 8.6, if you do it right and never trigger the creation of a string rep.

Running this in tkcon, the append turns the value into a string:

() 98 % set a [binary format H2 1]

() 99 % set b [binary format H2 1]

() 100 % ::tcl::unsupported::representation $a
value is a bytearray with a refcount of 2, object pointer at 0000000005665420, internal representation 000000000587B280:0000000005665240, string representation ""
() 101 % ::tcl::unsupported::representation $b
value is a bytearray with a refcount of 2, object pointer at 000000000564EEB0, internal representation 000000000587B4A0:00000000056590E0, string representation ""
() 102 % set x $a

() 103 % ::tcl::unsupported::representation $x
value is a bytearray with a refcount of 4, object pointer at 0000000005665420, internal representation 000000000587B280:0000000005665240, string representation ""
() 104 % append x $b

() 105 % ::tcl::unsupported::representation $x
value is a string with a refcount of 3, object pointer at 0000000005663F50, internal representation 0000000005896BA0:000000000564F030, string representation ""

this happens, because the bytearray has a string rep (due to Tkcon echoing the value) created. The append optimization only works for 'pure' bytearrays, e.g. bytearrays that do not have a string rep. This is similar to some optimizations for 'pure' lists.

So it works like this, preventing the shimmering result echo:

() 106 % set b [binary format H2 1]; puts "pure"
pure
() 107 % set a [binary format H2 1]; puts "pure"
pure
() 108 % set x $a; puts "pure"
pure
() 109 % ::tcl::unsupported::representation $a
value is a bytearray with a refcount of 3, object pointer at 0000000005658780, internal representation 000000000587B320:0000000005658CF0, no string representation
() 110 % ::tcl::unsupported::representation $b
value is a bytearray with a refcount of 2, object pointer at 000000000564ED60, internal representation 000000000587B500:0000000005658750, no string representation
() 111 % ::tcl::unsupported::representation $x
value is a bytearray with a refcount of 3, object pointer at 0000000005658780, internal representation 000000000587B320:0000000005658CF0, no string representation
() 112 % append x $b; puts "pure"
pure
() 113 % ::tcl::unsupported::representation $x
value is a bytearray with a refcount of 2, object pointer at 0000000005658690, internal representation 00000000058A5C60:0000000005658960, no string representation

Note the no string representation part.

How to turn a string into a bytearray

Just do a binary format:

set x [binary format a* $x]

Upvotes: 0

Related Questions