12

This question is a follow-up to How to use the text-processing tag effectively?

Recently, a question on text processing was tagged as . The question itself didn't mention any tool the OP needed to use/had problems using to achieve the task. Given that grep actually wasn't even suitable for the purpose, I interpreted this as the user being "not sure which tool to use" in the sense of the tag wiki - which I considered the consensus on the usage of these tags - in which case the usage guideline is to use the tag instead.

Since no other specific tool was asked for in the question text, and one of the tool-specific tags () was clearly misplaced, I felt the question did not fall in the "If your question is about multiple tools, include this tag and the tags for the other tools." use case, so I removed the tool-specific tags (in addition to the tag) from the question and replaced them with the generic tag.

This has prompted a discussion on the use of tags, in which one argument against this kind of re-tagging is that users looking (e.g.) for awk-based solutions will have a harder time finding this question and related answers. I tend to disagree with this view since that

  • would warrant putting all text-processing-tool-related tags onto such question, defeating the purpose of the five-tag-limit, and it
  • would mean applying tags in an attempt to characterize the (possible) answers, rather than the question.

Request for discussion

  • Should tool-specific tags be used on questions that do not explicitly ask for how to do this with a specific tool, or that do not describe problems arising from the (incorrect) use of a specific tool to perform the described text-processing task?
  • If the answer to the above is "no", does this warrant removing tool-specific tags on such questions in favor of the tag? Of course, asking the OP whether they unintentionally used an assortment of tool-specific tags when the text-processing tag would have been more appropriate is probably the first step here; my bad for not thinking about that.
  • What can be done to ensure visibility of questions tagged for generic text-processing tasks for future readers that look for solutions using a specific tool (such as )?

Procedure suggestion

Anyone can post their proposal for handling such situations in an answer. We may then choose to accept the highest-voted answer as community consensus (but maybe there is a better, established way on how to "decide" such issues?).

19
  • 1
    This is a great discussion to have; thank you for bringing it up!
    – Jeff Schaller Mod
    Sep 15, 2021 at 16:05
  • 1
    On a side note, I try to keep in mind that various viewers/answerers have different UNIX platforms available and sometimes those the sed/awk/etc toolsets vary across those platforms, so to me, a linux tag is helpful guidance from the author about the kind of tools they probably have available.
    – Jeff Schaller Mod
    Sep 15, 2021 at 16:06
  • 1
    Agreed. I find a tag for their specific shell very useful too so we know if a bash or zsh or other shell solution is appropriate. Many times some shell-specific construct is useful to find the files to operate on or to call the specific tool that does most of the actual processing of those files, etc.
    – Ed Morton
    Sep 15, 2021 at 16:28
  • 3
    When a user adds sed or awk tags to a question they're telling us the kind of tools they have available and think would be appropriate for solutions. Often we add/remove such tags that are appropriate/inappropriate while providing answers. So we get clues initially on how to answer and we get clues in future for how to find answers we want using specific tools. What is the benefit of removing those specific tags and replacing them all with a general text-processing tag? I see the benefit of adding text-processing of course, I just don't know the benefit of removing everything else.
    – Ed Morton
    Sep 15, 2021 at 17:04
  • 4
    Is this basically asking if tags should represent the question (only) or the answers (also)?
    – ilkkachu
    Sep 15, 2021 at 22:57
  • 1
    If someone posts a text-processing question, and answers are provided using awk, (pure) Bash, ed, Perl, sed and zsh; all with the asker using Linux, should the question be tagged with [text-processing][linux][awk][bash][ed][perl][sed][zsh]?
    – ilkkachu
    Sep 15, 2021 at 23:02
  • 3
    @JeffSchaller, the description for the linux tag explicitly says "If the question just happens to be in a Linux environment, please specify your Linux distribution in the body of your question, but do NOT use the /linux tag." That would appear to be at odds with the idea of using it to hint what tools they have. Plus the mere kernel doesn't tell anyway, e.g. an embedded system with mainly Busybox isn't exactly the same target as a desktop with the full set of GNU utilities.
    – ilkkachu
    Sep 15, 2021 at 23:07
  • 3
    As for following the hints the question gives about what tools are available. If a question is tagged with [sed] (only), or says something like "how would I do this with sed?", I'm not sure how much it'd be worth to use that as a basis to denounce answers using e.g. awk, since they're both standard tools and I suppose if one is available, then most likely another is available too? Or are there well-known systems that only provide one but not the other? (Same for the inverted case, and for other standard tools also.) ...
    – ilkkachu
    Sep 15, 2021 at 23:26
  • 3
    ... It would be different if they explicitly mention that they need the thing as part of a larger awk program (or whatever), or that they know the system they're on only has sed, but not awk. Also, there's the thing that a number of questions seem to ask about "doing that stuff in Bash", and taken at face value, that would mean both awk and sed (and join and paste and...) would be "out" for those. Which doesn't appear to be such a good idea. ~~~ [No, I don't have an answer. Just, thoughts.]
    – ilkkachu
    Sep 15, 2021 at 23:32
  • Yes, @ilkkachu, the Linux tag and I don't get along. It's my problem, and I'm having trouble getting over it :) I'm coming from a multi-OS background, and so I see value in having tags to filter the operating environment.
    – Jeff Schaller Mod
    Sep 16, 2021 at 11:29
  • 1
    @JeffSchaller, yep. But it would require somewhat a separate set of tags, something like "gnu-system" or "gnu-userspace" for any system with mostly the GNU command line tools (regardless of if it's RHEL, Debian, or some other OS, where they just have installed the GNU tools because they like them, and are open to installing more of them if needed.). Or "busybox-system"/"busybox-userspace" for the other similar case. (And then you fall into the trap where someone specifies "gnu system", but it turns out to rule away some non-GNU tool that's regardless commonly available on e.g. Linuxen...)
    – ilkkachu
    Sep 16, 2021 at 11:34
  • 1
    re. tagging answers, btw, on law.SE there seems to be somewhat of a custom of adding tag markers signifying the location to the answers, at least among some posters. Significant there, since legislation is different in different jurisdictions, not that I expect it'd help much with searching. See e.g. the answers to this Q.
    – ilkkachu
    Sep 18, 2021 at 13:49
  • Please use accurate edit comments.
    – Scott - Слава Україні
    Sep 22, 2021 at 6:03
  • 1
    @Scott Oh wow. I am the one prodding you. Seriously? The only thing you found was a bug or ill-implemented feature on the review queue of tag wiki edits. For your information, I performed two edits to the tag wiki. The first one in which I created it (which was then placed in the review queue for others to approve), and then somewhat later another one in which I removed ambiguous wording and a typo I oversaw in my very own text. If for whatever reason the edit queue mechanism doesn't preserve the edit history correctly, that is hardly my fault.
    – AdminBee
    Sep 23, 2021 at 7:41
  • 1
    Also, if you really want to play that game, I would be interested to know where the cosmetic part of this cleanup is to be found - I only see a simple rewording, and the place where you could have used e.g. inline code formatting to highlight the verbatim search pattern example (chunkbig10_1039 or eppix* - the latter being prone to misinterpretation as formatting tag btw.) and thereby actually improve the appearance, you have omitted.
    – AdminBee
    Sep 23, 2021 at 10:43

2 Answers 2

Reset to default
6

is that users looking (e.g.) for awk-based solutions will have a harder time finding this question and related answers

I have to challenge this because is a common yet misguided argument. If people search for awk based solution, they would include that in their search keywords and any useful search engine would scan the entire page and index all the content. If any answer mentions awk, it would include it, tag or not. Even SE own search engine (which depending the kind of search can be crappy or passable) is like this, which is why searching for "word" yields more results than [word].

Tags are meant to be consumed by potential answerers, searching is basically a secondary effect of the first. Which is supported by the excellent points that this other answer exposes on.

5
  • 3
    This. Tags don't make answers any more easy to find.
    – terdon Mod
    Sep 17, 2021 at 13:58
  • @Quasímodo law.se don't even need it, as long as the jurisdiction is mentioned. The New York one for example, would still be clear is about NY, by the first sentence. so does Finland, Brazil, Israel (last two uses titles instead/too). As you can see, tag isn't any different than mentioning it in the answer body, and don't provide any useful metadata unlike it does in questions.
    – Braiam
    Sep 19, 2021 at 13:14
  • @terdon Do you think we have a consensus, then? If so, how would we formulate it?
    – AdminBee
    Sep 24, 2021 at 15:38
  • @AdminBee well, deciding if there's a consensus or not is always hard. Among the people who have chosen to vote and/or answer here, there is a clear trend against using tool-specific tags when the question isn't about that tool but is that enough to call a site-wide consensus? That's always a tricky one. What we usually do is just point people to the meta discussion if they disagree and they can always post their own answer and see if it gets any traction.
    – terdon Mod
    Sep 24, 2021 at 15:43
  • @terdon Yeah, I guess you are right. Pointing people to this Q in case of objections will be the right thing to do.
    – AdminBee
    Sep 27, 2021 at 14:48
10

Should tool-specific tags be used on questions that do not explicitly ask for how to do this with a specific tool, or that do not describe problems arising from the (incorrect) use of a specific tool to perform the described text-processing task?

No, because

that would warrant putting all text-processing-tool-related tags onto such question, defeating the purpose of the five-tag-limit.

Does this warrant removing tool-specific tags on such questions in favor of the text-processing tag?

Yes, for the same reason.

What can be done to ensure visibility of questions tagged for generic text-processing tasks for future readers that look for solutions using a specific tool (such as awk)?

At the moment, nothing, I fear. If we could tag answers, probably that would be a solution. (At Codidact there is a proposal to allow tagging answers.)

Now, to be honest I think what you describe rarely occurs. Since most visitors of Unix & Linux are on a mostly POSIX system, I reckon they don't care much if the solution to their problem is Grep or Sed or Awk, so I see that as a smaller issue than tag flooding.

Some example questions for discussion:

  • Why does this Sed fail to do X and how can I do X?

    Both and fit — the first part of the question is clearly about Sed.

  • How can I do X? I tried this Grep but it fails.

    Given the wording, answers are allowed not to address the failed Grep. It is a question, but, is it a ?

    It depends on whether the task is doable with Grep such that one can point how the asker could fix his attempt. This case may be admittedly ambiguous; I'd not dwell on it if someone strongly disagreed.

  • Difference between -F , and BEGIN{FS=","} in Awk?

    There is no text processing task to solve. As an abstract question, it is only.

3
  • 1
    There was one such proposal here, too, but it was declined.
    – AdminBee
    Sep 16, 2021 at 10:54
  • 1
    There's also the thing that if you want to find solutions using awk to text-processing questions, you should be able to search for [text-processing] awk, i.e. "awk" as a keyword and not a tag. Not that I know if SE's search would still prioritize the tag too, but OTOH, some other search engine might not give the tags that much value.
    – ilkkachu
    Sep 16, 2021 at 11:39
  • 2
    The search engine allows for searching for [awk], awk, which is turned into [awk], "awk" (not converted into a tag) and code:"awk", and all of them can be combined with, for instance, is:q/is:a. IMO there is not much need for tagging answers, while a question I see is: (how much) do we expect users (especially new ones) to know about the search options they have? How much do we care?
    – fra-san
    Sep 16, 2021 at 12:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .