Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Launch HN: Evidence (YC S21) – Web framework for data analysts

news.ycombinator.com

Post reply on HN
Print view View original

61–70 of 94 posts

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#61

Post by newusertoday »

bayesian_horse wrote: post #58
newusertoday wrote: post #56

I have been doing this in emacs org mode for years i.e. reading from database and plotting the results along with documentation in single org file which is version controlled using git, how is this different? I can also export to html and share with others. something like this #+begin_src emacs-lisp (require 'ob-sqlite) #+end_src #+RESULTS: : ob-sqlite * Headline 1 * Headline 2 #+begin_src sqlite :db users.db :colnam…

This looks already way more complicated than "Evidence". OP wrote that people have been doing similar things for a long time. They are aiming to standardize and polish this approach.

Is that because of unfamiliarity with org syntax or something else? Current org approach of using code and data is already standardized and used in many organizations. So the value add is new approach by restricting it to markdown and only supporting sql and restricted set of components?

1. here is plot equivalent of evidence #+plot: set:"xlabel 'day_of_year '" set:"ylabel 'cum_vol'"

2. sql equivatent, note that org mode also supports other languages i.e. python/bash etc. that is the reason you need to specify sqlite here #+begin_src sqlite :db users.db :colnames yes select count() as total_calls, countif(timestamp_diff(current_timestamp(), created_date, day) ```summary select count() as total_calls, countif(timestamp_diff(current_timestamp(), created_date, day) 3. Document * Recent Call Volume Spikes # Recent Call Volume Spikes

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#62

Post by ricklamers »

This would have been a great chance to extend notebooks and make them the development environment for great looking reports (through a transform step from the completed notebook).

I think if you ask whether analysts enjoy coding in markdown vs in a notebook with SQL support you get a pretty unambiguous answer.

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#63

Post by krishvs »

bayesian_horse wrote: post #57

In my experience, most reporting has to be done as printable PDF or it doesn't exist. There doesn't seem to be a good story for paginated media yet. In the past I've used polyfills for CSS pagination features to generate modestly well formatted PDF documents. Then you can either print it out from the browser or use a headless browser to do the same. Also there are modules on NPM for Latex-style typesetting. This look…

Absolutely agree with this point. At the company I work for whenever we create reports its always done twice or thrice - a browser version, a pdf that is printable and an excel version for business users to download and play with.

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#64

Post by bayesian_horse »

newusertoday wrote: post #61
bayesian_horse wrote: post #58

Earlier quoted context omitted.

This looks already way more complicated than "Evidence". OP wrote that people have been doing similar things for a long time. They are aiming to standardize and polish this approach.

Is that because of unfamiliarity with org syntax or something else? Current org approach of using code and data is already standardized and used in many organizations. So the value add is new approach by restricting it to markdown and only supporting sql and restricted set of components? 1. here is plot equivalent of evidence #+plot: set:"xlabel 'day_of_year '" set:"ylabel 'cum_vol'" 2. sql equivatent, note that org…

Good luck getting non-emacs users to adopt emacs for this purpose. Of course you know that Emacs is the Best there is. It's just that ordinary people seem take so much convincing...

And still, your code has lots of "line noise", so I consider it uglier, but that's a personal preference.

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#65

Post by sails »

amcaskill wrote: post #31
sails wrote: post #16

> In our experience, the BI layer is the weakest part of the modern data stack. The BI layer has a poor developer experience, and decision makers don’t really like the outputs they get Totally agree, very interested in trying this out. FWIW I've tried and been frustrated by Looker, Metabase, PowerBI, Superset, Redash. I do think that while dbt does a great job with dimensional modelling, the BI layer is still require…

You have cut straight to the heart of a pretty interesting problem that we are still thinking through. You are exactly right, right now this is pure front-end. That's intentional, and there's a lot that we like about that approach, especially in light of the success of dbt. We think dbt fundamentally changes what is needed from a BI tool and that vendors who are maintaining really heavy built-in data transformation l…

>I think it would be very lightweight

Very interesting, that sounds great. I like the idea of keeping minimal business logic in the BI tool, but it is necessary.

And yes, lots of metric layer startups, taking advantage of that does make sense.

Will take a closer look, thanks!

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#66

Post by amcaskill »

saadatq wrote: post #59

> In our experience, the BI layer is the weakest part of the modern data stack. The BI layer has a poor developer experience, and decision makers don’t really like the outputs they get. It turns out, these two issues are closely related. The drag and drop experience is so slow and low-leverage that the only way to get all the content on the page is to push a lot of cognitive load onto the end user: global filters, dr…

Haha thank you! Unfortunately I think a lot of people are suffering under the status quo.

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#67

Post by amcaskill »

ricklamers wrote: post #62

This would have been a great chance to extend notebooks and make them the development environment for great looking reports (through a transform step from the completed notebook). I think if you ask whether analysts enjoy coding in markdown vs in a notebook with SQL support you get a pretty unambiguous answer.

Thanks for checking it out!

I agree that the developer experience of a notebook is great. We think we’re going to be able to replicate the parts of that dev experience that we really like.

We’re particularly interested in building some tooling that leverages the vscode notebook APIs, for example.

There are a lot of amazing tools for publishing & automating Jupyter notebooks as reports. All of these are worth a look:

google collab, databricks, mode, azure notebooks, deepnote, papermill, hex

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#68

Post by amcaskill »

bayesian_horse wrote: post #57

In my experience, most reporting has to be done as printable PDF or it doesn't exist. There doesn't seem to be a good story for paginated media yet. In the past I've used polyfills for CSS pagination features to generate modestly well formatted PDF documents. Then you can either print it out from the browser or use a headless browser to do the same. Also there are modules on NPM for Latex-style typesetting. This look…

Thanks for taking a look and for the kind words.

Absolutely, a great PDF export is on the roadmap, and raises a whole raft of tricky issues.

Just an aside, but one of the other features Sean and I really liked in R markdown was the ability to export a single standalone HTML file that you could just email to someone. I’m not sure if we’ll build that, but there are a bunch of cases where you want to send the “real” thing, without actually deploying anything (if that makes sense). It also saves you from the challenges of pagination.

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#69

Post by amcaskill »

swyx wrote: post #55

congrats on launching and building atop SvelteKit! Since you have some data processing inline with your Markdown, I am wondering if you explored using MdSvex? https://mdsvex.pngwn.io/ it would seem like it would be a pure win for you as you (your users) get reusable functions, on-document styles, local component state, and so on.

swyx!

Absolutely, we are using mdsvex, and we owe you a beer!

You rock.

Re: Launch HN: Evidence (YC S21) – Web framework for data analysts

#70

Post by amcaskill »

pm wrote: post #51

I can second Svelte and SvelteKit, despite the latter still not being at a proper 1.0 version. The developer experience was so smooth. I’ve been a fan of Vue for a while but Svelte and SvelteKit converted me almost instantly.

For real. The jump to svelte kit happened when this was just a hobby project, but I’m glad we did it.
Post reply on HN