Database support

[enjoysmath]

This will sound strange at first, but the usage's for this are manifold.

I'm looking into ArangoDB. Ran into trouble with Neo4j. ArangoDB has a JS driver so that's how it would connect into Quiver.

ArangoDB's AQL language supports regex matching, so for instance if you wanted F(A) to match G(B) as long as the substitution is common to the whole diagram, then that would handle that issue (having too many graphs, with all possible combinations of letters / variable strings). However, if the user starts to draw with F in mind, while matching a graph with G in the database, the user can stick with F and the substitution is applied transparently.

So every diagram can be stored in a graph in an arrango db (even if it has loops). Why do this? It would allow us to create a sort of gigantic Finite State Machine of high-level maths for one.

The user's drawing has a Category drop-down in which they can select what category they're working in, e.g. Ab. A functor button will allow the user to essentially relabel all object and arrow main labels L to F(L) if F is the given functor name.

I've thought about doing all this using a formal proof system but the code for that is voodoo to me. I've tried many ad hoc solutions that just don't meet the criteria you'd expect them to.

I do my work mostly in PyQt5. Here are some screenshots, actually of a previous attempt which involved Neo4j:
https://math.stackexchange.com/questions/2849246/category-theorists-would-you-use-a-software-tool-for-diagramming-chasing

What could you do with such a tool? Speed up learning mathematics (at your own pace) because the database would enable us to "gameify" category theory. "Fill in the missing homomorphism" e.g.

It could be used as a research tool. Think of it as a humanistic / not formal proof assistant. Everything is human checked, though there could be some ways of issuing an error via graph matching. If the logical error is present in the diagram when read by a human, then store that diagram with an error discription.

Commutative diagrams should highlight the faces of commuting squares, rectangles, etc in unique colors when you hover over them. This should all be done in 3D however, Quiver is a great start in this direction.

A user can step through a proof, with all uneeded info dimmed or hidden, all with clicks of a button. They can search for their diagram, or add it to the global database. Posted diagrams are rated just like any other site e.g. the MSE voting system.

All-in-all it's a tool to speed up learning about abstract mathematics, mainly maths which involves CDs. That is where the bottleneck lies. The best sources to learn from have been written for the past couple hundred years. It is however, our own learning inabilities that stop us from learning the material at a "quick enough" pace.

A tool like this could blow wide-open the doors of accademia to anyone with a PC or mobile phone. Why write a paper, when you can digitize your findings in a much more sophisticated tool than digital ink & paper (PDF / LaTeX documents). Or it can be used in conjunction with the old ways.

Discuss!

[varkor]

Regarding quiver specifically, I am very happy to discuss how to enable external tools to interact with quiver (e.g. exposing an internal representation for the diagrams, or possibly thinking about some sort of API). I also think interactive learning and proving using a visual editor is worth pursuing. However, I do not see this as a design goal of quiver itself, and I personally am unlikely to spend any time on these sorts of features. I think @AviCraimer is interested in similar ideas to you, so it might be worth starting a conversation with them.

[AviCraimer]

Hi, yes I am still interested in helping with this. Sorry I haven't been much in touch. I've just been busy with some family stuff. Looking forward to working on this!

[enjoysmath]

Hi, yes I am still interested in helping with this. Sorry I haven't been much in touch. I've just been busy with some family stuff. Looking forward to working on this!

How / when should we begin working together on this?

[enjoysmath]

@AviCraimer

[AviCraimer]

Maybe we should have a brainstorming / planning session sometime soon. I'm moving house this month so I can't commit to any coding work until February. I'm available to share ideas though.

[enjoysmath]

Awesome. I'm doing some reading on subgraph isomorphism algorithms as they apply to modern graph db's. IIRC, Neo4j doesn't even have a straightforward subgraph isomorphism search. So we should have some fun writing one of these in terms of breadth or depth-first search abilities of ArangoDB. I think graph search should be comprehensive in its options. For example, do we only match two arrows if they have the same hook/vee/split tail setting? That should be a checkbox. As I said before, this is a humanistic math project, so whatever we do doesn't have to be formally perfect and rigorous, as it will be a humans job in the end to verify correctness of some diagram or some relationship to another diagram. A brainstorm here seems like the right place for it. A harder problem to think about is do we try to autodetect variable names in a user's LaTeX label, or do we say that every string is a variable and so every string can be replaced by any other string in a morphism that describes a search matching. If you detect variables for the purpose of forming a regular expression, then what do you do with things like \lim etc which is everywhere in category theory. This is another reason why the search options should be comprehensive and detailed because maybe either way is sometimes preferable. I think the first goal should be just a simple search and display of an exactly matching graph (or whatever these papers I'm reading lead us), that is also a morphism of labels. That last part is quite important, you wouldn't want 10,000 copies of a commuting square to accomodate all latin and greek variable strings (with sub / super scripts, etc.). I believe these papers don't reallly address that way of doing it and instead they look for an injection of the label sets between the respective query graph and the graphs residing in the DB.

…

On Mon, Jan 4, 2021 at 11:27 AM Avi Craimer ***@***.***> wrote: Maybe we should have a brainstorming / planning session sometime soon. I'm moving house this month so I can't commit to any coding work until February. I'm available to share ideas though. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#63 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAMIF57IYXNFGG4TX53U7JLSYIJB3ANCNFSM4VQXSSGQ> .

[enjoysmath]

*injection = inclusion of the label sets On Mon, Jan 4, 2021 at 12:08 PM Luna Tuna <[email protected]> wrote:

…

Awesome. I'm doing some reading on subgraph isomorphism algorithms as they apply to modern graph db's. IIRC, Neo4j doesn't even have a straightforward subgraph isomorphism search. So we should have some fun writing one of these in terms of breadth or depth-first search abilities of ArangoDB. I think graph search should be comprehensive in its options. For example, do we only match two arrows if they have the same hook/vee/split tail setting? That should be a checkbox. As I said before, this is a humanistic math project, so whatever we do doesn't have to be formally perfect and rigorous, as it will be a humans job in the end to verify correctness of some diagram or some relationship to another diagram. A brainstorm here seems like the right place for it. A harder problem to think about is do we try to autodetect variable names in a user's LaTeX label, or do we say that every string is a variable and so every string can be replaced by any other string in a morphism that describes a search matching. If you detect variables for the purpose of forming a regular expression, then what do you do with things like \lim etc which is everywhere in category theory. This is another reason why the search options should be comprehensive and detailed because maybe either way is sometimes preferable. I think the first goal should be just a simple search and display of an exactly matching graph (or whatever these papers I'm reading lead us), that is also a morphism of labels. That last part is quite important, you wouldn't want 10,000 copies of a commuting square to accomodate all latin and greek variable strings (with sub / super scripts, etc.). I believe these papers don't reallly address that way of doing it and instead they look for an injection of the label sets between the respective query graph and the graphs residing in the DB. On Mon, Jan 4, 2021 at 11:27 AM Avi Craimer ***@***.***> wrote: > Maybe we should have a brainstorming / planning session sometime soon. > I'm moving house this month so I can't commit to any coding work until > February. I'm available to share ideas though. > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#63 (reply in thread)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAMIF57IYXNFGG4TX53U7JLSYIJB3ANCNFSM4VQXSSGQ> > . >

[enjoysmath]

https://community.neo4j.com/t/how-does-the-running-for-tiime-for-subgraph-isomorphism-search-scales-with-number-of-disconnected-components/31506

@AviCraimer

I'm back to evaluating Neo4j. ArangoDB didn't work out very well.

[enjoysmath]

const host = 'bolt://localhost'
const username = 'neo4j'
const password = 'lunamoona'

import { createRequire } from 'module';
const require = createRequire(import.meta.url);

const neo4j = require("neo4j-driver")

const driver = neo4j.driver(host, neo4j.auth.basic(username, password))
const session = driver.session()
const nodeLaTeX = 'A'

try {
    const result = await session.run(`
        MATCH (n:Object {LaTeX: $latex})
        RETURN n`, {latex: nodeLaTeX})
    const singleRecord = result.records[0]
    const node = singleRecord.get(0)
    console.log(node.properties.LaTeX)
}
finally {
    await session.close()
}

await driver.close()

@AviCraimer this is how easy Neo4j. I got this to run in a few minutes (compared with Arango, which took hours).

You'll need to install Neo4j Desktop and create a database named as above with a node with a LaTeX property with value "A". The code prints "A", successfully!

I'm in the process of learning Cypher (the query language of Neo4j), which is the string containing "MATCH" above.

[enjoysmath]

@AviCraimer
I'm not sure that JavaScript is the best language for running graph queries from. I think we should make a server-side Python / Flask application that communicates with Quiver and the Neo4j database (on the server).

[enjoysmath]

@AviCraimer

See this SO post on how to call a function on your flask app (server-side) from JS. I'm thinking we don't want to try and hack too much of Quiver, as it will have to be merged later with updates of Quiver. I'm guessing there's some way to embed Quiver into a subframe of the page, and have our own app parts (JS/HTML) running in other frames. We might add a button say to Quiver, but that's easy. According to a comment by @varkor in an issue thread here, we should subclass QuiverImportExport into QuiverImportExport.databaseService just like QuiverImportExport.base64. Althernatively we could somehow decode the base64 output, and that would be even more modular!

[enjoysmath]

https://stackoverflow.com/questions/59975596/connect-javascript-to-python-script-with-flask

[AviCraimer]

@enjoysmath,
Thanks for sharing all this. I won't have time to look at any of this in detail until after Jan 18th.

My first thought based on what you've posted so far is that there could be a natural division of labour here with you working more on the backend/database stuff and me working more on the frontend/UI. I want to make it easy to use create and embed arrow diagrams in a notebook-style webpage for interactive teaching materials and to capture/sketch ideas.

[enjoysmath]

@AviCraimer I like those ideas. It would be good for educators to use. To be honest, I'm starting to doubt my own abilities in completing this database support idea. However, I will try my best!

[enjoysmath]

Also, note that creating many databases is out-of-the-question and represents a major logistics / db management headache (I asked on the Neo4j forums). Therefore, I propose a single database for everything! Two nodes can be connected by a morphism if and only if they lie in the same Diagram. The Diagram node connects to all its node "children" via ":ContainsObject" (colon means Node/Edge label type). The :Object nodes will link to the Diagram as well via the inverse relationship ":InhabitsDiagram". The Diagram is then given a category assignment and also has a bidirectional arrow for ":AdmitsDiagram" / ":InhabitsCategory".

[enjoysmath]

So of course each diagram has an originalAuthor property and also a list of Editor. Also maybe we should store diagram edit history in the database? Let me know your thoughts on that :)

We most definitely want to be able to sequence diagrams in a proof in the database. This is random, but maybe we should have only one :Object named "A" in the whole entire database!

[enjoysmath]

Let me know when you would like to install Neo4j Desktop / Wingware (for a python IDE if you're new to Python). You can borrow one of my Wingware Pro license keys, which will give you the best debugging experience. I got a Flask app running in Wing, and you can step through it in the IDE!!! @AviCraimer

[enjoysmath]

@AviCraimer I've switched to a Django app and have read through the Neomodel documation. Here's where I'm at, defining the models for the graph database:

from neomodel import *

NAME_MAX_LENGTH = 100

class Object(StructuredNode):
    category = RelationshipTo('Category', 'IN', cardinality=One)
    morphisms = RelationshipTo('Object', 'MAPS_TO')
    name = StringProperty(max_length=NAME_MAX_LENGTH, required=True)
    

class Category(Object):
    objects = RelationshipTo('Object', 'CONTAINS')

Feel free to recommend any way of structuring things logically.

[enjoysmath]

Here is some progress:

That was constructed from python using a test file containing:

C = Category(name='C')
C.save()
x = Object(name='x')
x.save()
x.category.connect(C)
C.objects.connect(x)
y = Object(name='y')
y.save()
y.category.connect(C)
C.objects.connect(y)
f = x.morphisms.connect(y, {'name' : 'f'})
f.save()

As you can see, things are moving along. Neomodel is definitely the right choice for us. I will keep you posted here with future progress.

[enjoysmath]

@AviCraimer I have been working on getting a Django site hosted on heroku, it's at: quiver-database.herokuapp.com and I have the Neo4j database installed at enjoysmath.com. You can test the connection to the database with:

Clear database
Create test data

In that order. Note that it works! Took 2-3 days to get hosted, and get a dev environment setup. We can git push straight to heroku and the changes will appear on that site. The hosting is $8 / month though. We should slightly monetize to pay for a good Neo4j host eventually (such as GrapheneDB).

[enjoysmath]

I still have to get quiver (original) embedded in a Jinja2 template. I will work on that now, so much to do :)

[enjoysmath]

@AviCraimer It's a lucky coincidence because I forgot to google it before hand, but Heroku doesn't yet have Neo4j support (they only support a range of applications), however support for this as an easy Add-On is on the horizon it seems:

https://neo4j.com/news/neo4j-in-the-cloud-with-heroku/

So as soon as that comes out, we can host everything on heroku. Working with heroku has been top notch thus far.

[enjoysmath]

The only way to embed the quiver app is with iframe. I tried to integrate the file structure with Django, but that turns out way too difficult. This method however, worked almost immediately.

So we will have to edit parts of quiver to add in things such as "Save to Database" button. The methods I googled about for getting the iframe's URL as it changes did not work.

The above screenshot has quiver served from my local drive. All it takes is to put entire quiver folder inside of DjangoProj/static/. Then you can grab the index.html using an iframe and {% static quiver/src/index.html' %} as the link

[enjoysmath]

@AviCraimer I'm wondering now how you're doing the Front-End design because how I do things in Django depends on it. I've come up with a pipeline for UI dev. We first layout bootstrap templates using LayoutIt:

https://bit.ly/2LWN0N8

There's one I worked on. Django interacts nicely with bootstrap forms through a plugin called crispy forms.

Try going to Preview on that page, and resizing your window. You'll see that because we used Bootstrap (A CSS / JS library), it responds nicely to page size changes. So eventually we may be accessible from a phone screen.

[enjoysmath]

@AviCraimer

Quiver Database repo

The home page etc is not ready, but you can view this:

Rule Editor front-end demo

This is essentially the only way to do nice-looking, responsive, Django front-end templates. I.e. using Django + Bootstrap 3 + X-Editable.

Try clicking on an editable text, as you can see X-editable gives us this ability to edit text items in-place. It's not connected to the database yet. I will work on that next. Crispy forms with Django didn't work very well (had errors with it). Plus, I don't like how the frontend code is put on the python side (Crispy Form "Helpers").

So please learn some Bootstrap 3 and X-Editable. I used Layout-It to get a basic boilerplate template for that Rule Editor page:
Layout It.

---

The rule page essentially says, if you have a diagram drawn that matches the Key Diagram, then you can get to the Result Diagram logically. So when the user is drawing, we might populate a button widget as soon as a match is made. If the Rule Title is "Take Functor Image". Then that is what the button will read. The functor name / target category is supplied by the user per the rule editors options (Prompt The User) and the fact that a diagram is drawn containing in its text the variable "F", which is unassigned, so the prompt will read "F = ?". The other variables appearing the diagram are gotten from the key diagram (the bound variables). Those will get replaced by the variables actually used in the final user's drawing. Seamlessly!

[enjoysmath]

@AviCraimer you may have noticed when you go to change the category on the page, it outputs a weirdly formatted HTML error (403). I've gotten past that now! See: SO post.

I'm excited, now that we're past CSRF hell, I can be a lot more productive. I just need to follow the recipe that works.

[enjoysmath]

@AviCraimer It's better if I just make you a video of what I'm working on. So far, have gotten the Key Diagram Category field to update a Neo4j database with uniquely created Category nodes. We don't want a thousand copies of one category named "A category of left R-modules" for example. And unifying things some will help in the long run with applicatons of our tool.

Unlisted YT Video

[enjoysmath]

@AviCraimer I've put together a live demo for you.

Live Demo of the Site

It doesn't do any graph stuff quite yet. To demo it, click "Sign in" with these credentials:

TestEditor
fusion123

Then click "Rule editor". When you edit text, it's actually updating in the Neo4j database the corresponding node property.

---

Note that if the site loads slowly, that is only a one-time thing and only in the dev environment. Heroku runs code on a "dyno" and when say you press Sign In and it's spinning forever, that's the dyno starting up. I think they sleep them after 30 minutes of inactivity. In production this won't happen and the site is actually very fast. Once you wake up the dyno, you can see how fast it is on those DB writes.

[enjoysmath]

@AviCraimer I'm switching over the templates I have so far to Jinja2. With the new site structure, I have "Create New Diagram" pages embedded into a Rule Editor page, so that those Create New Diagram pages then open up a Blank Diagram once you have it named. I'm making Diagrams, Categories, DiagramRules all uniquely named within the database. This should encourage people to pick descriptive names. It's a while before I have it ready for you to clone from github though. Maybe 3 days. But I've come up with design to finished product pipeline (I've gone from design to nearly finished pages), so we should stick with it only adding 3rd party javascript or django plugins. Right now JS plugins are: Bootstrap 3 (not 4) for nicely scaling responsive layouts, X-editable (for in-place editing of text widgets), KaTeX as part of Quiver for rendering (so we should also use it for text in the rest of the site as well), and of course the quiver CD editor is embedded as an iframe. I tried twice breaking up the quiver files into a Django app hierarchy and failed. This is the simplest route.

We should share screens some time and I can show you what's happening in my dev setup and also how to set up your own. Don't be afraid to learn some Python, it's a lot easier than JS in my op.

[enjoysmath]

@AviCraimer

Here is the Latest Demo

Sign in with:

TestEditor
fusion123

How's it going? Are we still going to code together on this?

[enjoysmath]

Update: tried triggering events inside the iframes to no avail. So I tried making a django app out of quiver, and I finally got KaTeX to load, which I think was the main issue. Here's a screenshot:

This means we're going to replace all those JS buttons with our site's buttons outside of the render view.

This also means we don't have to worry about triggering events across iframe boundary. We can call javascript functions directly from the HTML template now.

Notice that we even have the URL with the diagram data string (see address bar of screenshot). We can use URL for now for passing data, but as the diagrams grow we have to keep in mind that URL length limits might get exceeded.

@varkor could you show me how to expose some functions out of the Toolbar class so that I can call them from index.html?

[enjoysmath]

Thanks @varkor. Also, I noticed that I can't put any extra html elements on the quiver page without shifting the grid and causing it to mess up when drawing. So I ask, is there a way to render the quiver screen to a div instead of the full viewport?

[varkor]

If you position the elements with position: absolute; or position: fixed;, I think it ought not to be a problem. On the other hand, if you want the elements to move with the canvas, you can add them to the element returned by the query document.querySelector(".container .canvas").

[enjoysmath]

@varkor the position: absolute worked, however I am unable to now click on the buttons, how can I "place them on top of quiver"?

Nevermind! z-index: 1 works in CSS of the bootstrap elements.

[enjoysmath]

If you position the elements with position: absolute; or position: fixed;, I think it ought not to be a problem. On the other hand, if you want the elements to move with the canvas, you can add them to the element returned by the query document.querySelector(".container .canvas").

@varkor I am using the bootstrap 3 CSS library. I've narrowed a bug down to something in bootstrap.css (attached).

bootstrap.zip

See how the grid is offset?

And the arrowhead end point is way offset from the cursor.

Any idea how I could narrow down what is causing this behavior?

[enjoysmath]

Fixed it with: changing "fixed" to "relative" in main.css:

.container {
    position: relative;
    width: 100%;
    height: 100%;
}

[enjoysmath]

@AviCraimer

Here's a video of the code succsesfully saving the diagram to the database:

https://youtu.be/bwnHyLQ4eoQ

[AviCraimer]

@enjoysmath

Looks like you've been making great progress. I've been moving into my new house this week so I haven't had time to follow this in detail. I will have some time Sunday afternoon. I'll take a look at it, and try to get the code running on my system.

[enjoysmath]

@AviCraimer The project was on the backburner for a long time. Now it's called DiagramChaseDatabase and it's a Django + Neo4j + Bootstrap + Quiver project. Are you still wanting to contribute?

https://www.youtube.com/watch?v=IU7_qV0FMRQ
https://www.youtube.com/watch?v=V1orXCXHJS4

[enjoysmath]

@AviCraimer I had to switch back to Django Template Library from Jinja2-based plugin. Ran into errrors that google couldn't solve which shouldn't have occured.

Anyhow, I have the site back in working order (not live but on my local dev machine) using DTL.

I had to switch from Jinja2 because I couldn't get a basic Crispy Form to work (Django plugin "Crispy Forms"), without erring out on some library's internal {% load x %} usage, which Jinja2 doesn't support, so not sure how things got crossed. None the less, I've got us back to the same state we were in that you can now see in the live demo, except with the DTL being used.

When we connect. We need to exchange emails her in a code-talker like way to keep them secret, then we'll delete posts.

Once we've done that, you need to install TeamViewer (so I can RDP into your machine to help you get the dev env setup).

Once we do that, we exchange connection details over Skype. Let's connect over skype first, We'll need a private chat channel for certain parts of the dev.

I need to install:
Python 3.7
WingWare IDE Pro
A list of required python files in requirements.txt in the project's root dir.
I'll show you how to do that using Python's pip command. It's like npm of Node.js.

Heroku (what we will first host our live site at) looks for a requirements.txt or a setup.py. In our case we just use requirements.txt because it's simple: It's just a newline separated list of known packages with optional "== version number" on each line.

So for instance:
PyQt5 == 5.6.8

You do pip install -r requirements.txt in the same directory that is the project root i.e. where manage.py resides. manage.py has subcommands runserver 8000 which will start the Django server hosting your site (locally) when you do:

python manage.py runserver 8000

You will also sometimes need to do python manage.py makemigrations for staging changes to your database model (in models.py of each subfolder that is an independent django sub-app (like "polls", "registration", "user_profile", "blog")).

So any time you make edits to a models.py file, run the above makemigrations command.

Then run pythong manage.py migrate to update your actual database (which is a SQLite 3 file in the project's root directory, but which according to the Django docs we don't want to use in production and it will be switched over to a PostgreSQL database on Heroku - which is a task to set up still on my to-do list).

The code in settings.py has a switch:

if ON_HEROKU:
    // setup for PostgreSQL when live on Heroku
else:
   // setup for SLQLite3 on local dev machine

You can code settings.py, and when you are debugging code in WingWare, changes to settings.py or anything non-model related (even .js or .css files in static) are usually reflected immediatly in your browser without restarting the debugger / django server's process. You just have to save the file, so if it shows a "*" in the tab name, then hit Ctrl-S to save it.

If the change isn't reflected, then hover over tab titles in your IDE's an make sure the directories are correct. Also you might want to do a Chrome > "..." icon > Clear Browsing Data > Use default settings for common dev files. To clear out your browser's cache of .css / .js or image data that has actually changed but the browser didn't detect it.

There's lots to learn but I can get you up to speed more quickly than going at it alone and reading tutorials. So you'll be where I'm at in a a week of coding together.

I'm a big copy / paste coder. If it's already coded perfectly to suite my needs, I just copy / paste code and if it's slightly off I adjust settings or go to other examples, etc. It has been hectic trying to code login / logout / account signup views as a new person to Django. Each tutorial is slightly not just what I need, so I have to keep jumping around... Maybe you would be good at coding the signup / login(out) / password recovery system. Remember we have a $8/month linux host we can use if we need an email system, but maybe Heroku too has a cheap option.

So there are two servers:
https://quiver-database.herokuapp.com/

Has an exact copy of the code that we push with:

git push heroku

Once you have a remote added to git for heroku.

The second server is:
enjoysmath.com

(which hosts our graph database). But for you to deploy to to the live site you'll need the secret keys and to know where I store them. So we need a private Skype conversation going on.

I will show you how I make changes to both servers. If second logon costs money, I'll lend you my login for Heroku. I think though I should create a user for the linux server that has some admin priveleges like you can start / stop neo4j.

[enjoysmath]

@AviCraimer There have been some changes to the README.md page of the QuiverDatabase repo.

https://github.com/enjoysmath/QuiverDatabase

It now gives instructions on setting up your environment. Let's move future discussion over to the issues section of that repo, or someone can show me how to set up the Discussions tab there.

[enjoysmath]

@AviCraimer Here's a screenshot of where the code is at:

I'm implementing Rule Search as we speak. This allows the user to match an input diagram with the key (or result) of a DiagramRule. Then hitting "Apply" will take the user to the new diagram that is the old diagram with the found rule applied to it. In theory.

Having some issues with regexes and Neo4j querying, so made a Neo4j community forum post about it here:

https://community.neo4j.com/t/regex-matching-text-hom-does-not-work/33295

Are you still interested in working on this together?

[enjoysmath]

@AviCraimer

I got the Rule Search feature + Rule Application (with Variable Substitution) features to work properly, and here is a video:

https://youtu.be/kKJySXo9BDA

[enjoysmath]

I think I'm about to re-open this project from the backburner. I thought there must be more rigorous ways of doing this CD database, but what I have so far is pretty far along, so I'm going to stick with it and try to get it fully working.

[enjoysmath]

@AviCraimer I'm getting back into Quiver Database code. Are you interested in collaborating on it?

I'm thinking we fork Quiver and make it as close as possible to original Quiver except set / get graph ability and anything else needed complete the design. I did a lot of hacking of JS directly into my copy of Quiver. But that makes updating from this official repository not so easy.

One thing I would like however is better mobile support. The tool buttons are not fitting on my screen and when you zoom out on the iPhone the buttons will then be within the screen, but too small to click on comfortably. Other than that the mobile support is amazing. You can draw diagrams with your finger in such a way that it's much easier than using a mouse.

So for example, to decouple the code more than what it is currently, the search results would not be a popup within Quiver JS code, but a panel rendered in the Django template itself, outside of Quiver's iframe. See above screenshot for what I mean.