D. Moonfire @dmoonfire@octodon.social

My proof of concept of #AuthorIntrusion worked! I got it to find 9.5k echo words in my 600k word serial in 40 seconds. Since this appears to be a viable option, that means I have to switch from randomly working to project management and start issues in Gitlab for what is a 1.0.0.

Of course, it looks like I'm going to have to create a website for the NuGet packages, a lot of usability things, hooking to Atom via language services, but I see a possible path forward.

I got echo detection working on a 150k novel going through 44 chapters in 7 s. Only taking about 0.1 s per file.

There are some usability and testing things to make life easier, but this means I hit my first milestone.

(And it found a couple problems in my current novel.)

I'm very, very happy right now.

Finished monthly accounting for Broken Typewriter Press. It wasn't a good month, maybe next time.

I will want to automate more of it, mainly because I'm lazy. Maybe PayPal API for sending out payments or figure out how in the world I can scrape the reports from the various ebook vendors since they won't provide me APIs.

... I also wished I sold more so I can start a magazine or periodical.

Last one. Why I'm doing this. I got into a discussion with a coworker about it.

Grammar checkers are great, but they don't look for stylistic issues. So reducing adverbs, detecting echo words ("she jumped on the jumper, jumping on his knee to shatter the bone"), my personal love affair with gerunds, or pointing out when I accidentally used present tense in my novel.

That's what AI is mainly for. To catch those things that an editor would catch.

With the previous iterations of AI, I switched to write-novel/update-code look that was really useful. Of course, this was back when I wrote an entire damn text editor with customized keyword highlighting and auto-corrections.

I failed because I was basically trying to write Scrivner.

The more recent iteration is to write a focused tool (ala `git`), and then integrated it into other tools (e.g., Atom, Emacs, VS Code) to handle everything else.

This also means there is only a few steps left until I can see if this proof-of-concept works. I need a "check" command that will run the echo detection across the entire project and outputs the results in something GCC/Emacs/Atom can use. At that point, I'll have the minimum needed to use Make or a build command and get useful, specific results.

Using C#'s XSLT function overrides and XElement has really simplified the code. I'm also happy that I've gotten rid of a big "reimplemented wheel" that the last iteration had (implementing CSS generically) since I get lost in the stupid details of getting those "right".

And since you can pull in XSLT functions through packages, that means I can easily write a package that provides "soundex()" or "stem()" or "part-of-speech()" to look for more complex items.

Another productive lunch hour on #AuthorIntrusion while working on echo word detection.

I got it to correctly detect the same word used within a given range. With these plugin settings:

- scope: content
select: //token[@l>3]
within: 3
warning: 1
error: 2

And the file:

one two one three four one

It produces a warning on the first and third "one", an error on the second "one".

For xpath, I'll probably change it to `//token[length()>3]` because it's prettier.

I really like when complex plumbing code suddenly reduces down to three lines because I finally got it working.

That happened tonight as I finished the tests for paragraph and token splitting. Then I could use it to start getting echo detection working.

That excites me because the first milestone is getting closer. I'm really hoping that it will also be performing once I throw something big at it. Right now, I'm going to say Flight of the Scions because it is 130k words.

I think my danger is if I stop before I have something concrete to use while writing a novel, I'll lose track and feel like I have to start over.

Though, I don't know if I have time to integrate it with Atom which would be ideal.

The rest of the house was sleeping, so I got two hours of coding on Author Intrusion. Took a bit for the infrastructure, but I got Markdown paragraph splitting implemented.

The xpath worked out fairly smoothly, which helped a lot. Once I get token splitting, an easy task, I did be on track for getting echo word detection by end of week.

Then I'll get timing on my 600k word serial to find performance issues and call the results milestone one.

I love this chapter because of the cat woman. Also because she is the only healer I have planned on writing for at least ten books.

My view on magical healing came from a book on rpg design with a good essay about how healing and Mana potions are crutches to keep players going but in novels, the downtime is just as important as the action scenes, that last bit comes Dwight V Swain's book.

Flight of the Scions 12: Virsian #Action #YoungAdult #Patreon #OneDollarPatreon
https://fedran.com/flight-of-the-scions/chapter-12/

Surprised by Falkin's offer of a better room, Kanéko gets a chance to enjoy a hot bath. Then she is surprised by another dalpre, a cat woman named Virsian.

Writing via TDD is occasionally slow, but it also feels like I'm trying to write a solid base for the rest of the code. Since I've failed so often to write #AuthorIntrusion, this approach "feels" better to me.

Now, I don't do BDD, I do a lot of TDD testing on SRPd internals.

The plugins will have also a "select", so looking at the first word of every paragraph in a file would be:

scope: content
select: //para/word[position()=1]

To look at previous chapters, you would just change scope from "content" to "project".

I'm still planning on making the XSLT functions DI, so I can later write a function like `soundex()` that would let me compare like-sounding words that are too close to each other.

select: //para/soundex(word)