gaorlov / sqhshare_graphs Goto Github PK

This is pretty much the opposite of another request, so maybe we make a flag for it?

The idea is this: when we plot a graph (x,y), we currently wrap the SQL with SELECT x,y FROM (...). This makes the JSON returned as efficient as possible. This request from @billhowe is the opposite:

If some flag is set, make the "title" text of the point ("tooltip"?) contain all the non-plotted columns of the tuple. So given a relation

When moused-over, the graph would highlight x=5 and y=3 as it currently does, and the title text of the point (that should pop up near the mouse cursor) would say:

x : 5
y : 3
a : "a"
b : "b"

or something like that.

It would be good to include string columns as a valid choice for bar charts.

Queries can get long, so it may not always be appropriate to display the underlying query on the bag view. That way, the view can double as a dashboard.

The initial load of the main "DataSets" page is a little poky and returns all datasets. Maybe switch to the search API and return just the user's owned datasets by default, then allow simple keyword search.

The embed options under "Share" aren't HTML, they're URLs. Is it possible to put code you can just drop in a webpage? E.g., <iframe src='...'/> or whatever better code you'd recommend?

If the "Share" link is open and the graph is regenerated, the URL isn't updated. Maybe just close that panel when the graph is changed?

On the full dataset view something is happening with the display, perhaps when there are lots of columns. See attached.

Right now, if I view the dataset "[[email protected]].[NSAF avg SpC with avg NSAF]" in SQLShare, I see the data. But if I click the dataset in SQLShare-Graphs, it appears to take the sql_code from Emma's dataset and run that---which leads to a "Select permission denied" error.

This code pattern generally will not work -- it is VERY common for someone to make a view public but not the underlying data. In fact, the 'table_blah' table created for every view 'blah' is usually private.

So when I click a dataset [[email protected]].[dataset], the derived SQL should be either "SELECT * FROM [[email protected]].[dataset]" or "SELECT TOP 100 * FROM [[email protected]].[dataset]". This probably fixes the "weird code" you mentioned above.

I suppose this would be a feature request.

On a simple X-Y plot, it would be great if you could select what is displayed.

Here is an example
direct link to SQLShare graph: http://goo.gl/2Iy2iu
There are three columns, the first two I want to plot x-y and have column 3 display when hovering over point.

If I try to edit a query to have something like

...

where some_column like '%somestring%'

I get a 404.

I had a "full dataset" query (Tokyo_3_merged.csv) return the attached error message. I suspect it was just because there are way too many records -- bad choice on my part, but we may want to think about how to fail here.

Not sure this is the right diagnosis

Check the line chart at the left here:
http://sqlshare-graphs.herokuapp.com/bags/billhowe@washington_edu/Tokyo

Timestamp is the x-axis, but seems to show up as a color on the chart, and perhaps as a result the scale is goofy.