gojq -nr '["\u0000"] | @tsv'

[pkoppstein]

Compare:

$ jq -nr '["\u0000"] | @tsv'
\0

and the output produced by gojq:

$ gojq -nr '["\u0000"] | @tsv'
�
# In the line immediately above, a single literal NUL should appear.

In 2015, there was discussion about what jq should do with NUL and
even though it may not be documented in the manual, the decision
to present NUL as \0 by @TSV was part of the discussion at
jqlang/jq#759

[itchyny]

Please let me know if any application that recognize \0 for NUL characters in tsv or csv. As far as I opened the output by Numbers, the gojq output is correctly recognized and \0 is recognized as \0.

[pkoppstein]

That's not quite the right question. The point is that a TSV file should be processable by tools that are NOT well-disposed to NULs:

$ /usr/bin/awk --version
awk version 20200816
$ /usr/bin/awk  -F\\t '{print $2}' null.tsv
b

$ cat null.tsv
a	b
c�	d
$ 

[itchyny]

So, there is no way to deal with NUL characters properly, and we should not use it in the input for the filters. Rather than making it indistinguishable against "\\\0", I prefer to keep it as it is to notify the user the wrong place in the input data.

[itchyny]

Hmm, I'd rather reject them by errors.

[pkoppstein]

@itchyny wrote:

there is no way to deal with NUL characters properly

But, when generating a TSV file, special characters (notably tabs, linefeeds, and returns) must be handled sensibly, and I believe, the same logic applies to NULs as well. Agreed, there is no "TSV standard" that mandates allowing raw NULs in a TSV file, but (a) there are plenty of applications which recognize the two-character sequences '\t', '\r', '\r' and '\0' as stand-ins; and (b) in the present context, jq itself establishes a time-tested and reasonable standard.

[itchyny]

Okay, I noticed that the escaped character is escaped in @tsv so that's fine. I'm worried about the behavior of @csv and @sh that they yield the same string against different strings (["\u0000", "\\0"] | @tsv,@csv,@sh,map(explode)). But after a few minute research, it seems that dealing with the null characters in CSV and shells are very difficult. After all, most people are happy with the tool just work against their real data, and few people are care about the behavior against the null characters.