💾 Archived View for gemi.dev › gemini-mailing-list › 001016.gmi captured on 2024-05-12 at 16:21:20. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2023-12-28)
-=-=-=-=-=-=-
I'm interested in opinions on syntax highlighting for preformatted text. Gemini has a lot of potential for distraction-free developer reference (I just discovered that godocs.io works in Gemini!) and syntax highlighting would help. In Markdown, you can add hints to preformatted text blocks like so: ```html <p>This is HTML</p> ``` This tells the client that the preformatted block is HTML, so the client can present the code in whatever way makes the most sense. Usually this is syntax highlighting, but I could imagine some clients adding collapsable code blocks based on tags, braces, and/or indentation depending on the language. The Gemtext spec would seem to have no problem with this, as the first line is considered "alt text" and it is up to the client to determine both what should be done with it and how it should affect the presentation. But I wanted to put this out there to get a broader perspective. I've started on a patch for adding this to amfora, since the chroma library (also used by Hugo) can do syntax highlighting with ANSI escape codes, for use in terminals. It works, but it's got a couple of rendering bugs at the moment. Discussion and screenshot here: https://github.com/makeworld-the-better-one/amfora/issues/252
mntn@mntn.xyz writes: > I'm interested in opinions on syntax highlighting for preformatted > text. Gemini has a lot of potential for distraction-free developer > reference (I just discovered that godocs.io works in Gemini!) and > syntax highlighting would help. > > [...] > > The Gemtext spec would seem to have no problem with this, as the first > line is considered "alt text" and it is up to the client to determine > both what should be done with it and how it should affect the > presentation. But I wanted to put this out there to get a broader > perspective. There was a thread on this topic some months back: gemini://gemi.dev/gemini-mailing-list/messages/005732.gmi Alexis.
> On Sep 12, 2021, at 11:14 AM, mntn@mntn.xyz wrote: > > I'm interested in opinions on syntax highlighting for preformatted text. Gemini has a lot of potential for distraction-free developer reference (I just discovered that godocs.io works in Gemini!) and syntax highlighting would help. > > In Markdown, you can add hints to preformatted text blocks like so: > > ```html > <p>This is HTML</p> > ``` > > This tells the client that the preformatted block is HTML, so the client can present the code in whatever way makes the most sense. Usually this is syntax highlighting, but I could imagine some clients adding collapsable code blocks based on tags, braces, and/or indentation depending on the language. > > The Gemtext spec would seem to have no problem with this, as the first line is considered "alt text" and it is up to the client to determine both what should be done with it and how it should affect the presentation. But I wanted to put this out there to get a broader perspective. Sounds like a neat idea. I can imagine _someone_ wouldn’t like it on the grounds that it’s on the noisy side color-wise, but a disabling toggle switch sounds like the sort of thing a buried preference would be for.
Hi mntn,
I have been wondering about the use of alttext details for formatting.
In the next couple of months I shall be experimenting with Parsing
Expression Grammars. The idea is to build PEGs, using content from
manpages and cheatsheets. I plan on doing it in the (regex and
multidimensional array friendly) Lisp, TXR.
Hopefully it may be possible to identify when a command is being issued
and deduce the format of the outputted code.
Of course, providing a switch 'sh' could open up things.
Even more so providing something with precision of software stacks may
prove even more fruitful
``` guix environment --ad-hoc python2-numpy python-2.7 -- python
print("Hey, I wonder: Is this more interesting than using Orgmode?")
```
Some ideas for a tonk client:
Thanks, I read over it and the deeper linked conversations. It looks like people have been considering this from a number of angles. It does seem like a lot of people were suggesting changes to the spec; I don't want to add anything new! Gemini is fine as it is. I am confident that the best option is just to use the name of the language as the alt text. Alt text of "javascript" is fairly clear as to what follows in the preformatted text block--so it still works as alt text--and a syntax highlighter could use that hint for parsing the text and highlighting the code. A properly designed syntax highlighter will give up and fail quietly/safely if it can't parse the code for whatever reason, say if it's actually a poem titled "javascript" instead of code. Clients may want to ignore anything following a space or punctuation, so alt text could be even more descriptive, such as "javascript: excerpt from example.js" or "HTML example." This would seem to be compatible with the spec as it stands, it doesn't munge the alt text, and it doesn't add any new requirements for clients. September 12, 2021 5:57 PM, "Alexis" <flexibeast@gmail.com> wrote: > mntn@mntn.xyz writes: > >> I'm interested in opinions on syntax highlighting for > preformatted >> text. Gemini has a lot of potential for distraction-free > developer >> reference (I just discovered that godocs.io works in Gemini!) > and >> syntax highlighting would help. >> >> [...] >> >> The Gemtext spec would seem to have no problem with this, as the > first >> line is considered "alt text" and it is up to the client to > determine >> both what should be done with it and how it should affect the >> presentation. But I wanted to put this out there to get a > broader >> perspective. > > There was a thread on this topic some months back: > > gemini://gemi.dev/gemini-mailing-list/messages/005732.gmi > > Alexis.
On Mon, 13 Sept 2021 at 16:18, mntn <mntn@mntn.xyz> wrote: > I am confident that the best option is just to use the name of the language as the alt text. Alt text of "javascript" is fairly clear as to what follows in the preformatted text block--so it still works as alt text--and a syntax highlighter could use that hint for parsing the text and highlighting the code. A properly designed syntax highlighter will give up and fail quietly/safely if it can't parse the code for whatever reason, say if it's actually a poem titled "javascript" instead of code. Personally I think using the MIME type would be most sensible, it's what's already done when a page is received, so it'd make development easier for clients. (I'm including non-""official"" MIME types. The ""official"" coverage of source code files is a bit lacking) > Clients may want to ignore anything following a space or punctuation, so alt text could be even more descriptive, such as "javascript: excerpt from example.js" or "HTML example." Big +1 to this :) These two examples given here don't seem to be useful, if the client is adding syntax highlighting then it can probably generate an alt-text saying "HTML source code" by itself. Stating that it's an example or something is highly likely to be done in the normal text anyways. However, I imagine making a screen-reader read code would be absolute hell! So it would be extremely useful for this to be a thing, so the code can be roughly described in a screen-reader safe way. -Oliver Simmons (GoodClover)
> Personally I think using the MIME type would be most sensible, it's > what's already done when a page is received, so it'd make development > easier for clients. > (I'm including non-""official"" MIME types. The ""official"" coverage > of source code files is a bit lacking) You have a good point. I wonder "why not both" -- a client could make a best effort to detect the language using several methods. For the example listed, "js" may also be appropriate alt text for JavaScript. Again, if it can't be parsed as the detected language, it wouldn't be highlighted. > However, I imagine making a screen-reader read code would be absolute hell! > So it would be extremely useful for this to be a thing, so the code > can be roughly described in a screen-reader safe way. That's exactly what I was thinking. I think this would both benefit users with screen readers (so they know what's coming) and people like myself who find it easier to parse code with some syntax highlighting in place. I can also imagine a highly advanced screen reader that might be able to parse code and present some kind of intelligent navigation. (Next/previous statement or expression, enter or exit nested code, etc.) I don't believe this exists today, but I can imagine it!
On Sun, 12 Sep 2021 18:14:07 +0000 mntn@mntn.xyz wrote: > The Gemtext spec would seem to have no problem with this, as the > first line is considered "alt text" and it is up to the client to > determine both what should be done with it and how it should affect > the presentation. But I wanted to put this out there to get a broader > perspective. That's not what alt text is. It's for screen readers. A bunch of people are way into ASCII art so alt text is needed. They are image descriptions. I recommend the following instead: Just name the language on the line before the <pre> block. ie.: lua: ``` local function whatever() end ``` This way you aren't making an accessibility feature worse by trying to overload it to be some generic metadata container.
> On Sep 19, 2021, at 1:43 AM, raingloom <raingloom@riseup.net> wrote: > > On Sun, 12 Sep 2021 18:14:07 +0000 > mntn@mntn.xyz wrote: > >> The Gemtext spec would seem to have no problem with this, as the >> first line is considered "alt text" and it is up to the client to >> determine both what should be done with it and how it should affect >> the presentation. But I wanted to put this out there to get a broader >> perspective. > > That's not what alt text is. It's for screen readers. A bunch of people > are way into ASCII art so alt text is needed. They are image > descriptions. > > I recommend the following instead: > Just name the language on the line before the <pre> block. > ie.: > > lua: > ``` > local function whatever() end > ``` > > This way you aren't making an accessibility feature worse by trying to > overload it to be some generic metadata container. Someone floated the idea of using the leading ``` line for alt text and the trailing ``` line for language colorization (or maybe the reverse). It sounded like a good idea to me, but the current maintainer of the Gemini spec (not unreasonably) has a try-it-first-and-report-back attitude to these sorts of things and, as far as I can tell, nobody’s modified a Gemini browser to try this sort of thing out. I use the leading ``` for both types of preformatted blocks and I’d be happy to change my capsule to use this sort of bifurcated thing, but I wouldn’t know what to try out first. My hunch is that moving the language-colorization text to the final position is the winning play, but that’s the option that goes directly counter to preexisting (Markdown) practice/habit/muscle memory, so it’s liable to be the sort of thing that’s screwed up more often (people typing colorizer hints where alt text should go).
Nathan Galt <mailinglists@ngalt.com> writes: >> On Sep 19, 2021, at 1:43 AM, raingloom <raingloom@riseup.net> wrote: >> >> On Sun, 12 Sep 2021 18:14:07 +0000 >> mntn@mntn.xyz wrote: >> >>> The Gemtext spec would seem to have no problem with this, as the >>> first line is considered "alt text" and it is up to the client to >>> determine both what should be done with it and how it should affect >>> the presentation. But I wanted to put this out there to get a broader >>> perspective. >> >> That's not what alt text is. It's for screen readers. A bunch of people >> are way into ASCII art so alt text is needed. They are image >> descriptions. >> >> I recommend the following instead: >> Just name the language on the line before the <pre> block. >> ie.: >> >> lua: >> ``` >> local function whatever() end >> ``` >> >> This way you aren't making an accessibility feature worse by trying to >> overload it to be some generic metadata container. > > Someone floated the idea of using the leading ``` line for alt text > and the trailing ``` line for language colorization (or maybe the > reverse). It sounded like a good idea to me, but the current > maintainer of the Gemini spec (not unreasonably) has a > try-it-first-and-report-back attitude to these sorts of things and, as > far as I can tell, nobody’s modified a Gemini browser to try this sort > of thing out. Using the trailing ``` has a ugly drawback: we have to read all the code block before being able to colorize it. Gemini has a really nice property of needing only the first few bytes of each line to decide its "type", there's no going back and forward like it is for markdown (where link references can be declared after the link itself) or HTML. This allows a really cool property: text/gemini files can be *streamed* on the fly and easily parsed on the fly. For instance: a irc-like application can have a page that never stop loading where all the message "appear" as soon as someone sends them, and a second page where the users can send text. That said, regarding the try-and-report part, if someone really wants to see how this could work, I suggest try hacking something on top of elpher: Emacs is really cool for this kind of things and already has syntax highlighting for various programming languages. Cheers,
Omar Polo <op@omarpolo.com> writes: > Using the trailing ``` has a ugly drawback: we have to read all the > code > block before being able to colorize it. Gemini has a really nice > property of needing only the first few bytes of each line to decide its > "type", there's no going back and forward like it is for markdown > (where > link references can be declared after the link itself) or HTML. And having had to deal with parsing Markdown - due to Markdown libraries tending to assume that you want HTML output, and not making a parse tree available - i can confirm that this is a nice property indeed. Fwiw, as someone with disabilities (albeit not vision-related), and who has watched accessibility affordances on the Web get (to put it politely) short shrift over the decades, i tend to agree with raingloom: > That's not what alt text is. It's for screen readers. A bunch of people > are way into ASCII art so alt text is needed. They are image > descriptions. > > I recommend the following instead: > Just name the language on the line before the <pre> block. > ie.: > > lua: > ``` > local function whatever() end > ``` > > This way you aren't making an accessibility feature worse by trying to > overload it to be some generic metadata container. Alexis.
---