My current setup is a postgres db and jupyter instance running on docker. Data visualization is easy enough with seaborn/matplotlib/pandas plotting.
I would love to see a better tool for this but its hard to beat running your own. There is no vendor lock in, no subscription fee, no uploading large datasets, no privacy issues on private data, no "gotcha" moment when the tool can't produce quite the visualization you're looking for.
You seem to have a good setup, but if you are looking for a solution to share easily between people, have a look at Metabase.
It works pretty well, if your data lives in a DB hosted somewhere and you only want to plot data that can be derived from SQL queries (potentially from tables produced by nightly batch process), and want to allow different people to look at the data via dashboard.
I used to do all plotting in Pandas etc locally and share them if required. But with Metabase it is easier to share plots, which are also always fresh.
RAWGraphs seems to have a different target audience. Judging by the sponsors, it seems to aim at people who keep their data in Excel and then need a nice plot for a publication or something. So it might be useful as a prototyping tool for scientists who barely code and keep the data they collect in Excel spreadsheets.
I second this. I've deployed Metabase once and almost immediately requests for data started coming out of the woodwork, from folks who previously had no knowledge of what kind of information was available or how to ask for it. Visibility is called that for a reason.
It's also really easy to self-host. Download the .jar. Whack it on a machine with a wireguard VPN and a route to the DB or somewhere suitable in the cloud. Create a read-only credential. Letsencrypt it & bless it with a DNS name, ideally with some internal-only IP address. Invite your boss to view some live stats on what's happening with the live product. Wait a day or so for them to request that some other underling also gets access or that some Metabase query gain some formal role in a business process. Also watch them have the best week of their manager life, with every kind of stat at their fingertips.
I really wanted to like this, I've tried it several times but always end up with either having to pre-process my data to fit whatever vis I'm trying to create or just simply not being able to do it.
I recommend https://observablehq.com, once you get a hang of it and have a couple of templates setup for your workflows it's similar effort of this and you have a lot more flexibility.
I think this is by design. The idea is you bring your data in the expected format (using spreadsheets, or pivot tables, or observable, or pandas), and RAWGraphs will make one of a few common modern visualisations. The sort of stuff you would hope Excel/etc. will have out of the box someday.
Observable is definitely more of a power house :) Out of curiosity, do you have any templates you'd be willing to share?
I built a whole startup, GetBulb (still kind of up at https://getbulb.com) that pretty much did exactly this, but aimed more exclusively at Excel/spreadsheet users. This was around 2014. We received some investment but the market seemed to want consultancy work for custom visualisation over a set of templates that needed users to have some judgement on what template best suited their data.
I would consider pandas more of an alternative to spreadsheets. Both are used to create/massage/reorganize data. RAWGraphs is more like an alternative to seaborn? Albeit much more limited but much easier to use. The "missing link" marketing is a little odd, since it's essentially just a visualization library (you don't have to use it with spreadsheets if you don't want to)--but that's just marketing.
This tool uses the "drag and drop columns into the graph editor" pattern, which I've seen in a few other places too - Kibana Lens most recently: https://www.elastic.co/kibana/kibana-lens
I have to admit this interface doesn't really fit the way I think about charts. I'm interested in alternative UI designs for this kind of charting tool that don't use drag-and-drop - anyone seen any good ones?
Tableau follows this drag-and-drop pattern, but in a much more explicit way than Kibana, with the ability to drop columns onto slots with (more-or-less) predictable effects. It's immensely polished - and priced accordingly.
I'm actually not sure there's a better UI pattern for the nontechnical user. If you want to visualize both the slots and the options at once to give people a sense of what they can do with their current schema... you don't just want a dropdown for each slot that's going to forget your state, move all over the place, etc.
Maybe that's my problem: I'm a very technical user, and I find the drag thing baffling. Perhaps if I had less depth of knowledge I'd find it more intuitive?
I've always loved this UI. I think of visualisations as data dimension bound to visual dimensions (eg x position, y position, z position, colour, size, group, etc). So dragging the data dimension to the corresponding visual dimension afforded by the visualization maps perfectly with my mental model. How do you think of visualisations? Do you have an example of a UI that better matches your mental model?
One really really cool option I've seen is this research project called Aprtus: http://aprt.us/ . It creates this cool two-way binding between the properties of the visual elements and your data; supports recursive rendering/composition; etc. It's a little hard to explain, but the video in their homepage is worth a watch. That's probably my most ideal visualisations UX; easy to use, whilst still making it easy to use code to make complicated visualisations.
Long time lurker here, just created an account since my friend had the same issue as you, so I started working on https://github.com/wodend/pysimpleplotter. It is still in development, but please give it a try and let me know if it better suits your use-case, keep in mind this was specifically designed to work with CSV/TSV output from scientific instruments. We plan on supporting exporting charts with styles for specific journals, and I will add better error handling when I get time.
I’ve seen pyDatView: https://github.com/ebranlard/pyDatView which I use for time-series csv files generally. Might be some good inspiration for you too!
I'm one of the developers of Lens, and I agree that it's not the right fit for every problem. The way I categorize the different types of interfaces in my mind is a spectrum from:
* "shelf based" drag & drop like Lens, Tableau and most spreadsheets
* Query-based, where you type a query using SQL or another grammar and choose from a template
* High level code like ggplot, Vega or Observable's new Plot
* Low level code like D3
Since you're a developer, maybe you'd benefit from looking at one of the high level visualization grammars.
Power bi and Tableau offer a number of ways to transform data including python and R. There are a ton of plug-ins out there for custom d3 graphs or even using matplotlib right inside power bi. Check out "the charticulator."
I've also been using Google data studio and it is great and offers tons of way to connect to data and refresh for free!
I think that's a really good tool for someone would like to visualize their data with better way.
For me, the UI it's quite good and intuitive.
And generated charts also beautiful.
This works really nicely with our data munging tool. Just copy your data from Easy Data Transform (after any filtering, joining, cleaning etc) and paste it in. Choose the axes. Graph.
Now this is a "s/Facebook/Hacker News is tracking my life" coincidence. Just today I was looking to convert a CSV into a graph and came upon this exact site
I would love to see a better tool for this but its hard to beat running your own. There is no vendor lock in, no subscription fee, no uploading large datasets, no privacy issues on private data, no "gotcha" moment when the tool can't produce quite the visualization you're looking for.