text stringlengths 0 1.82k |
|---|
**Brian Ketelsen:** Oh, I know what you're talking about - the golang data library page. It's at mjhall.org/golang-data-science-libraries. It's a huge page full of data science libraries in Go and we all just kind of stopped and said, "Wow! There's 500 links on this page." I was thinking there were two or three data li... |
**Daniel Whitenack:** Yeah, and something I've seen is... Even still you talk to a lot of data people and they talk about Python or R, and you hear about people working in those languages, but the newer companies that I've interacted with or contracted with, they are adopting Go and other modern languages more readily,... |
**Erik St. Martin:** Are there any specific Go packages or frameworks that you're currently using? |
**Daniel Whitenack:** Sure, yeah. I mentioned Pachyderm, I am using that currently, and have had some great experience with that. I also very much like the idea that you can think in terms of these data pipelines and think in terms of piping data from one container to another. This is very useful, because let's say I w... |
Generally a lot of what I'm writing is in Go, but I like that flexibility. I also like the reproducibility of it, which is a pretty huge problem in data science right now. So the fact that you can commit your input data, your output data at every stage of the pipeline is very useful and powerful, I think. Also I've bee... |
For example, I recently did a k nearest neighbours sort of thing. In the spirit of the Go proverb, "A little copying is better than a little dependency", I basically just stole some of the Euclidean distance functions and that sort of thing out of Gonum and was able to throw together exactly the k nearest neighbours th... |
**Brian Ketelsen:** We did a project recently, last year, where I needed to do some data processing, and I found it really easy to take Python libraries and port them over to Go for data science type things. I can't remember what I needed to do, it was one of those numeric-type things, but it really was not difficult a... |
**Daniel Whitenack:** Yeah, and I think part of the mindset of like... In Go sometimes it's easier... Something I've seen - maybe you guys can comment on this - is doing... In Go sometimes it's just faster to write a for loop than it is to import a certain package to do some relatively simple thing. And I think it appl... |
**Brian Ketelsen:** Right. |
**Erik St. Martin:** Matt Holt is in our backchannel, which is GoTimeFM channel on Slack, the Gophers Chat. He said that he implemented k nearest neighbour in Go for his machine learning class. He said, "Great experience, totally worth it. Just inline it." |
**Daniel Whitenack:** Yeah, I totally agree. |
**Erik St. Martin:** To your point, I think that that's a good pattern when looking at importing dependencies in general, right? If you need one function - especially if it's small - do you need to import that giant library? All of it has implications too, so I guess it's always something to evaluate whenever you pull ... |
**Daniel Whitenack:** Yeah. And there are a good number of packages related to data science, including machine learning packages. I recently saw a data frame package called Gota that's in development, which will be definitely very nice for exploratory analysis, I think. I would be sad if doing data science in Go just f... |
**Brian Ketelsen:** So what kind of advice do you have for somebody who wants to get started with playing in data science in Go? Do you have resources or advice, or places you like to send people? |
**Daniel Whitenack:** Sure, sure. First of all, like I said, I think doing data science in Go should be a little bit different, so I think definitely starting out I would say that those same resources that other people starting out in Go find useful, like Peter's resources or Dave Cheney's, and kind of getting into tha... |
Then a lot of what I've learned has been trial and error; importing this package and trying that, writing my own custom stuff... Hopefully that's changing. A lot of data scientists come into Python and try things in Jupyter notebooks to try to figure out interactively how does this work. One of the things I've been wor... |
Then of course there is the list of things that have already been done in Go, like InfluxDB and Pachyderm and most sorts of things. Generally there is Getting Started sections to those projects. In Pachyderm they have a great Getting Started example where they do word count with simple grep and awk commands. That's kin... |
**Carlisia Thompson:** Daniel, we shouldn't be surprised that you're going to talk about Go And Data Science at the upcoming GopherCon, but can you give us a little teaser about the talk? Who should be looking forward to being there and seeing that talk? |
**Daniel Whitenack:** Sure, sure. So it's still in the works, but the teaser I'll give is that from start to finish basically I'm gonna do data science, and do it hopefully in a distributed way, all with Go. So starting out saying like "This is a problem we wanna solve. We're gonna explore the data with Go tools, and t... |
Then also I'll definitely provide some call-to-action to the audience to start doing some data sciency things with Go and point them to some packages and some projects that they can contribute to. |
**Carlisia Thompson:** So is it going to be like a live demo? |
**Daniel Whitenack:** Yeah, I mean it will be a combination between some code that I'll show on the screen, and then depending on how things work out, I'm hoping to show some type of live demo. |
**Brian Ketelsen:** I say you gotta go Kelsey Hightower or go home. |
**Daniel Whitenack:** Exactly. It's a scary proposition, but it could be a lot of fun. |
**Brian Ketelsen:** Just live code on the screen. |
**Erik St. Martin:** If you have slides, you have to deploy VMs that do this stuff from the slides. |
**Daniel Whitenack:** Sure, sure. Actually, I'll point to some Go notebooks and you can do it yourself in the notebook. |
**Carlisia Thompson:** Very cool. |
**Erik St. Martin:** So speaking of Go notebook, there's the Jupyter, right? And there's now a Go library for that? |
**Daniel Whitenack:** Yes, so there's a working kernel for Jupyter, it's called Gopher Notes. I started this back in, I guess it was January. It's functional now, so you can use it to make Go notebooks. There's definitely open issues and things to work on, so this is one of those things that I think hopefully some peop... |
**Erik St. Martin:** I think we're running very short on time, but for anybody who's not aware, can you give a brief rundown of what Jupyter is? |
**Daniel Whitenack:** Sure, yeah. There's actually a whole ecosystem of Jupyter projects right now. The main project is called Jupyter Notebooks, and what you can do is you can start a Jupyter Notebook server on your computer, and then if you go to the browser, you can start a Python notebook, and there's a lot of othe... |
**Erik St. Martin:** Excellent. We've just about exhausted our time here, but before we kind of transition into closing out the show, is there anything that you wanted a chance to share with everybody that you have not got a chance to? |
**Daniel Whitenack:** I think a lot of what we've talked about is definitely what I wanted to talk about. I would encourage people out there - there are a good number of people doing data science in Go out there, even if they're not the most visible in the data science community... So let' s say, all of you Go engineer... |
**Erik St. Martin:** The way we typically wrap this thing up is we have our \#FreeSoftwareFriday where all of us go around and thank some projects or contributors - or both - that are making our lives easier currently or in the past. With that being said - Brian, who do you wanna thank? |
**Brian Ketelsen:** I know our rule is no more than one, but I have two - forgive me in advance. |
**Erik St. Martin:** I did three one week, it's cool. \[laughter\] |
**Brian Ketelsen:** Okay, good. So the first is ngrok from Alan Shreve. My god, there isn't a day that I don't go by without using ngrok. It's just the best tool on the planet for sharing something that's running on your machine with people that are somewhere else. If you haven't used it, I think it's at ngrok.io, but ... |
**Erik St. Martin:** Carlisia, how about you? |
**Carlisia Thompson:** I'm going to mention Jupyter Notebook and the Go kernel. Daniel did a great job describing Jupyter Notebook, but I still wanna mention it because it is not just for data science, and I'm gonna tell you why. I came across this commercial tool that is a notebook, and my mind was blown. This is the ... |
**Brian Ketelsen:** I'm sold. |
**Erik St. Martin:** I kind of wanna look at it just for normal note taking. |
**Daniel Whitenack:** Yeah, and thanks to Carlisia for going through the setup and giving me feedback on the different issues as well. I definitely appreciate it and I'm glad people are starting to use it. |
**Carlisia Thompson:** Yeah, I did run into just a super minor hiccup, and I opened an issue. Daniel was so quick, he was instantaneously responding to what I was having trouble with. I think he'll do the same for everybody. |
**Erik St. Martin:** Uh-oh, she's setting a precedent. |
**Brian Ketelsen:** There's pressure... \[laughter\] On next week's show we're gonna talk about open source pressure... \[laughter\] |
**Erik St. Martin:** ...and how we can add more to the contributor. \[laughter\] |
**Carlisia Thompson:** Setting expectations... |
**Erik St. Martin:** We also like to ask our guests too if there's anybody they kind of wanna thank, any projects they'd like to bring highlights to... |
**Daniel Whitenack:** Yeah, I wanted to mention Vim Go because I've started using it recently. I know a lot of people use it out there, but I'm kind of a recent convert. I've just found it amazing, and I think it's... Once I kind of got in the process of using it and writing Go with Vim Go, I think it's improved my dev... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.