RE: Read a fixed length of input each time

[Tom Gray]

The usual way to do such things is with "glue" programs.
Maybe xxd will do the trick.
Google "man xxd"
Tom

-----Original Message-----
From: bug-gawk <bug-gawk-bounces+tom_gray=keysight.com@gnu.org> On Behalf Of 
Peng Yu
Sent: Tuesday, June 23, 2020 10:17 AM
To: Andrew J. Schorr <aschorr@telemetry-investments.com>
Cc: Bug Gawk List <bug-gawk@gnu.org>
Subject: Re: Read a fixed length of input each time

[EXTERNAL]

I don't think it can be done with one-liner in python. I am not sure about 
Perl, but even it can be done with a one-liner, its syntax can be too 
complicated than necessary.

Even bin2uint() can be made in an extension, the fixed-width parsing still 
needs to be supported by core gawk.

I think the gist of gawk is to treat input as stream and broken it into records 
and fields. Any input that can be broken into records and fields, in theory, 
could be processed by this model of gawk. But records and fields may not need 
to be denoted by deliminators. They can be deliminated by fixed length.

Thus, the text input restriction is not an essential limitation of awk. It 
could be that awk may have two modes, one for text input and the other for 
binary input. I don't think there would be any conflict between the two modes.

On 6/23/20, Andrew J. Schorr <aschorr@telemetry-investments.com> wrote:
> On Tue, Jun 23, 2020 at 11:45:32AM -0500, Peng Yu wrote:
>> For example, I may do the revert processing to get back the integers 
>> from the binary stream. For this kind of fixed binary stream, I feel 
>> the awk syntax would be better than writing a much longer Go file.
>>
>> For example, if gawk would be able to split fields based on fixed 
>> field width and a fixed number of NF. Then, I may simply write awk 
>> code like the following (suppose that there is a bin2uint() function).
>> It would be must shorter than the equivalent Go code to convert such 
>> binary stream to integers.
>>
>> $ awk -e '{ for(i=1;i<=NF;++i) { print bin2unit($i) }'
>
> To accomplish that, you'd need to write an extension library that 
> allowed you to specify the number of bytes in a record and the number 
> of bytes in each field. I'm not aware of any other way of achieving 
> that, but it should be pretty easy using the extension API.  And then 
> you'd have to write the bin2uint function, which I guess can be done 
> using the ord() function in the ordchr extension. But it still seems 
> like this is probably much easier in other languages such as Perl or 
> Python (or perhaps Go; I'm less familiar with that one). It's best to 
> use the right tool for a given task; AWK is designed for text 
> manipulation, but other languages are better for handling binary data.
>
> Regards,
> Andy
>


--
Regards,
Peng