alixedi / gitdash Goto Github PK

Seriously, this one is pretty crap. No numbers to add. Nothing exciting.

I showed this to my wife and she didn't know that she could drag the labels around.

I think the concept of pivot tables can also be better explained - by demonstration.

There are tools for this. For instance: http://bootstraptour.com/

As easy as pasting comma separated values in an input/Text box while in DataView Mode and pressing the button "AnalyzeThis" to see the visualizations. All Client End.

Example:
http://davidguttman.github.io/react-pivot/

Client Side Data can also be achieved by github's new feature of Drag and Drop any file easily to your repo and BAM! , you can analyze your data with a click of button!
Checkout how easy it is now to upload a file. An amateur with little knowledge of github interface can do it:
https://github.com/blog/2105-upload-files-to-your-repositories

We should have a clear guide for developers on how to contribute to the project. This should detail how to set-up jekyll for local development as well as any guidelines wrt pull requests.

I was sitting the other day drawing stuff with Abbas and I drew this:

Like? If yes, I need your help in digitising it.

So I have used a pretty small font overall to save on screen area.

The select boxes however are still fat.

We need this.

We should have support for sharing as PNG.

http://stackoverflow.com/questions/3975499/convert-svg-to-image-jpeg-png-etc-in-the-browser

At this moment, if you share a URL, it will not set up the pivot properly - specifically, the rows, columns and value drop areas will not get populated like they are supposed to.

Drilldowns are interesting. They might work as a filter (?)

For instance, consider this data:

Category, Thing
Animal, Cat
Animal, Dog,
Fruit, Apple
Fruit, Orange
Animal, Horse

Pivoting by category, we get 2 bar charts - side by side - One of them titled "Animal", the other titled "Fruit". If we click on "Animal", it naturally should mean that we would like more on "Animal" and so in the next screen, you basically only see the data that has "Animal" as a category.

The following is a good-ish example although, if you look at the code, the data is laid out hierarchically - which in my opinion is not necessary.

http://jsfiddle.net/gh/get/jquery/1.7.2/highslide-software/highcharts.com/tree/master/samples/highcharts/drilldown/basic/

So, right now if you look at our example data, you will notice 3 parties called "Democratic". This is because of leading and trailing whitespaces.

The solution should be simple - make sure each cell of the CSV goes through a trim().

We should allow people to put in a Google Analytics ID in the _conf.yml file.

I really think that out Dripl design will do for the moment.

The original pivottable had filters: https://jsfiddle.net/nicolaskruchten/w86bgq9o/

Myself and @alirazabhayani did not include filters in Dripl because we did not know how to represent them in the querystring.

Now, I have an interesting idea: What if we only do filters for classifier fields. I define classifiers as following:

is_classifier = lambda data: len(set(data)) < len(data)/20

IPE: For data with 100 records, a classifier is a field that has less than (for instance) 20 unique values.

This means that in our example POTUS data, fields titled "Party" and "Home State" are classifiers -http://bit.ly/1QWf4IG and http://bit.ly/1RqGIQa.

There is absolutely no need to have filters or indeed any pivot-table functionality for non-classifiers. What is the point of filtering a given president (only) from the dataset? E.g. http://bit.ly/20M3vfs

So in the step when we tell the user to filter Took office [Months], the onboarding popup disappears and never re-appears.

So in charts, there is no need for:

• Table
• Table bar chart
• Row heatmap
• Column heatmap
• Line chart

Bring back pie chart.

In aggregation functions, get rid of:

• All percentages
• Both binomials
• intSum
• sumOverSum
• etc.

Without tests, any and every change can potentially break the thing.

http://orizens.com/wp/topics/first-steps-in-setting-up-travis-ci-to-your-javascript-project/

This is an opportunity for self-promotion of sorts.

What should we put in there that makes sense?

If you try and drag a label off Rows or Columns when it's the only label in the area, but instead of dropping it in some other area, drop it back in the same area you started from, you end up with the label as well as the placeholder.

At the moment, we are referencing pivot.js and jquery.parse.js locally.

Both of these libraries are available via CDN. We should use it.

• We need to write a "How to contribute" guide @alirazabhayani had to waste time in setting up the branches, jekyll etc. This should not be the case. I am on it.
• The importance of pivottable library that we use I was chatting with @osamakhn about this. Our pivot library is an asset. It's the only thing that works every single time. We should keep it as our technical debt. Only to be replaced when the rest of the stuff is up to scratch.
• Importance of having competitors Found another one. Its not the same but not too different either. It is super-important to have these projects because it provides reassurance that whatever we are trying to do is important enough for other people as well.
• Miscellaneous Myself and @osamakhn also chatted about onboarding and README etc. I am going to add the findings of our discussions to the respective tickets.
• Unrelated but interesting I did a survey on FB asking devs the reasons they do not participate in FOSS. Most of them apparently do not have time.

It would be really nice to have images showing every single step and what needs to be done.

If we have a date column called Birthday e.g. 04 Jan 1989, we should expand it into 3 columns: Birthday Year, Birthday Month and Birthday itself.

We can ask our user which columns to expand and by how many using _config.yml.

This should allow us to do queries like - How many *** were born in 1989?

Cant have enough of these! Lets have a few in the footer:

https://ghbtns.com/

We really should have a pie chart.

• What do we want to get out of this exercise? An MVP: Same pivot table. GitHub backend. Awesome documentation, great onboarding, filters.
• What is the broader goal of the project? We think we have made the best instant dashboard on the web. We want people to use it. We want to build a community around it. We want to be able to improve it in the future.
• How can we make it happen? Spare some of your spare time for picking up issues, we can also plan a 1-day hackathon to make some quick progress if we feel it may be useful.
• How do we give credit to founders? We can have a section in the README that mentions the founders for the project.

Before we have filters, we need a filter UI.

See this for ideas: https://jsfiddle.net/nicolaskruchten/gq6w05j8/