Hacker News Reader

Comment by ajarmst | Hacker News Reader

I don't get it. Why would I do diagrams in ascii when I can use any one of dozens of good programs and export the result via DVI or PDF. I guess it must might be nice if I want to read the document on a text console, but I don't see that use case coming up nearly often enough to justify learning yet another version of markdown and commit to doing diagrams with ascii art.

View on HN

Because using ASCII makes the documents more portable, more cross-platform.

There is not need to learn another version of Markdown, the diagrams are just plain ASCII and Markdeep supports the Markdown you know and love, plus more.

I think it is great, I love it!

The "have ASCII diagrams converted to images" thing is getting ridiculous. No, I don't want to type a border around my diagram. Text is completely and utterly unsuitable for this, I don't want to have every change require me to reformat everything.

(And frankly, the diff thing is nonsense, too. Diff produces an unreadable mess on ASCII art. I'd rather see the diff of some XML that shows me nodes added, removed, properties changed.)

> Text is completely and utterly unsuitable for this, I don't want to have every change require me to reformat everything.

You need a better editor, that's all. Emacs has picture-mode and artist-mode; no doubt vim has something too, although a quick search didn't help.

I actually like the idea, for simple diagrams at least. No need to keep separate image resources anywhere. No need to even have a separate image processing program. Want to change that diagram? Well, it's right there in the document itself -- just change it. Done. And the diagram is legible both pre- and post-rendering of the document.

Exactly. There are lots of situations where it's worth trading up some polish for simpler tools and formats. Especially simpler formats.

There is lots of prior art extracting diagrams from textual drawing but AFAIK all required global conformance to some accepted "syntax". Markdeep's(?) idea of local beautification while maintaining the global fixed-width layout is brilliant as you're not limited to particular structures and anything unsupported degrades gracefully! (If this idea has no name, let's call it "2D ligatures"?)

We could have a dozen different tools based on this idea, each supporting a somewhat different dialect, yet any diagram will be reasonably readable under any of them — which is exactly the quality you want when excavating a 40-year-old README of unknown origin.

It's the simplest thing that could possibly work. Except one: using line drawing characters directly in the source (and requiring no tool at all, just view in fixed-width font). Let's try:

  *                                                                             .                 *
  *    0       3                          P ●              Eye ╱         ▶     ╱                  *
  *     ●───────●      +y                    ╲                〈)          ╲   ╱  Reflection       *
  *  1 ╱│    2 ╱│       ▲                     ╲                ╲           ╲ ▶                    *
  *   ●───────● │       │                v0    ╲       v3           ────────●────────             *
  *   │ │4    │ │7      │                  ●────╲─────●                                           *
  *   │ ●─────│─●       +─────▶ +x        ╱      ◀ X   ╲          ◜⎺◝◀────────        ○           *
  *   │╱      │╱       ╱                 ╱        ○     ╲        ( ╱ ) Refraction    ╱ ╲          *
  *   ●───────●       ▶                 ╱                ╲        ◟⎽◞               ╱   ╲         *
  *  5       6      +z              v1 ●──────────────────● v2    │                ○─────○        *
  *                                                               ▼                               *

Hmm, not as good but mostly OK (though font dependent), except for circles. Disappointingly I found no matching parts for bigger than 2x2 circle. It's easy (technically, not sure about politically) to add a few more but it's impossible to support smooth circles of arbitrary radiuses. (OTOH, neither does Markdeep.) Perhaps the best of both worlds would be writing unicode (perhaps using some ascii->unicode helper at typing time) and further enhancing it with this kind of tool...

Version control. If I add an ASCII art diagram to a text document, I can git diff it, and see that the ASCII art diagram was added. If my documentation is is in PDF, I can keep it in version control, but I lose a lot of the advantages that version control provides because PDF is a binary format.

What diagram tool uses PDFs? PDFs are a build artifact like any other binary.

But if you create the pdf using a script (e.g. latex tikz) then this is in your version control instead.

Yes, once we have more intelligent diffs. Line by line isn't enough anymore.

I agree, it's kind of a one-time affair too because if you ever need to add or change a word in a node, you're fcked and have to edit ascii to change borders and lines and what not. There are way better ways to spend your time than to draw an ascii diagram.

http://asciiflow.com/

If there's a way to edit boxes/lines once you draw them to address kgen's point, I've never been able to figure it out in asciiflow.

I've only been using it for a few seconds, so I don't know if I'm off-base here, but isn't that what the third-to-last icon at the top does (box with diagonal arrow)?

Thanks, you're right. I was thinking of the old version. http://stable.ascii-flow.appspot.com/#Draw

Well, the only worst think would be to have to launch Visio, Powerpoint or whatever and then export and then re-embed. Especially if you want to fix something done by someone else so you only have the image and not the binary... Also, versioning.

1. You need n copies of the tool where n is the number of people in your team.

2. Version control for binaries sucks.

3. Graphical tools can be very awkward when aligning stuff. MS Word, for example, is a pain for making diagrams in.

LaTeX and DOT are not graphical tools.

There are literally dozens of examples of text-based markup tools that are cross-platform, version-controllable, and can be rendered in a terminal if you're into that, and which don't require me to waste time drawing ridiculous ASCII art borders to denote logical relations.

Reply on news.ycombinator.com