Disqus - Latest Comments for cobaltchris

Re: Adding New Fields to Your JSON Decoder

cobaltchris — Wed, 05 Apr 2017 18:23:00 -0000

I'm not sure what was going on when I tried this, but I have this working now. Thanks for the article and explanation, in particular on the maybe decoder.

Re: Adding New Fields to Your JSON Decoder

cobaltchris — Wed, 05 Apr 2017 00:24:23 -0000

I'm wondering how to do this with a custom decoder for the optional case? I've been trying to get it to work, but not understanding how to construct my custom decoder so that it can work as the 2nd argument to the optional (but also still work properly with a regular required field that isn't producing a Maybe). e.g. I want something like:


activity = 
    decode Activity
        |> required "total" myStatsDecoder
        |> optional "sleep" (map Just myStatsDecoder) Nothing

myStatsDecoder : Decoder Stats
myStatsDecoder 
    decode Stats
        |> required "hours" int

So, my JSON will for sure have the "total" key, but may or may not have the "sleep" key. Both of those would have a Stats value.

Re: Creating Your First Elm App: From Authentication to Calling an API (Part 2)

cobaltchris — Thu, 27 Oct 2016 22:54:34 -0000

Great tutorial, thanks so much for providing this!

Re: Sublime Text 2 for Ruby

cobaltchris — Sat, 23 Jun 2012 02:14:15 -0000

Good writeup! A few of us at HotelTonight are using ST2 as well. Some plugins I like include "Alignment", "Ruby 1.9 Hash Converter", "SideBarGit", and SublimeREPL (now if could just get this to work for Rails console).

For the ctags stuff, I have a small shell script that updates the tags, and run that via cron twice a day. Script looks like:

ctags -R --exclude=.git --exclude=log --exclude=tmp --exclude=autotest --exclude=db -f .tags * ~/.rvm/gems/`rvm current`

So, that indexes a Rails codebase (run this from inside a Rails root dir), and includes the gems in your RVM gemset (if you're into that). I'd be interested to hear how others keep their ctags file up to date.

Re: How I manage 40 people remotely

cobaltchris — Wed, 13 Jun 2012 01:18:45 -0000

At HotelTonight, about 75% of our employees work in our San Francisco office, however, I am partially remote. I'm the CTO and run engineering and part of product. I live in Eugene, OR (similar choice for quality of life, etc.), but spend about half my time in SF these days for high bandwidth communication and collaboration with the team. I fly back and forth every other week or so, and of course stay in hotels via HotelTonight (shameless plug :)

This is not without its woes, and at the pace we move, and now with just over 50 employees, it's tough. Our developers are heavily involved with product and we don't organize product the more typical ways, which has worked great so far, but we're always questioning and willing to change to adapt to company growth and so on.

Tools help, and as many others have mentioned, we've had great luck with HipChat. The dev team makes great use of it, and our stellar support team make extremely heavy use of it (integrating with various other services and parts of our system).

We use Pivotal Tracker to manage dev work, Google docs, Dropbox, JungleDisk, and are currently experimenting with the Q&A portion of Yammer.

As Ryan mentioned, TriNet is a huge help, and recommended. We have employees in several states, as well as some in the UK as well.

I've worked remotely a lot over the years, and find that the success depends significantly on your team, product, and stage of the company. When I was at Adobe, they were very supportive of remote, but not all individual teams supported it (some due to particular managers, others due to constraints such as high cost hardware testing lab needs). Other companies I've been involved with have been fully distributed, with nobody in the same office. A lot will depend on the communication quality (and comfort) of fellow team mates, possibly language barriers, and then significant time zone differences definitely impact things. My own experience is that it works best if individual product teams aren't more than a couple time zones away, otherwise it just kills velocity (the flip side is that you can get 24/7 ops or support coverage without people having to pull graveyard shifts). It's also key that the company buys off on it in general, and usually it's best to have multiple remote people, vs. just say one or two who can more easily be forgotten.

Finally, I'd say it's certainly worth trying if you have need, but just note that it's not guaranteed to work. With the right tools, team members who are more than open to it, and great communication, you can be successful.

Re: BlueForest - A dark color scheme for IntelliJ IDEA ← Decodified

cobaltchris — Thu, 16 Jun 2011 16:29:37 -0000

This is outstanding, great work! I've been too lazy to properly recreate the "Argonaut" theme from TextMate in RubyMine/IntelliJ, but BlueFlower is very similar, and now that I've used it for a couple days I'm just loving it. Thanks, much appreciated!

Re: Hoptoad and Javascript, Sitting in a Tree

cobaltchris — Wed, 04 Aug 2010 18:46:33 -0000

We too are seeing these same rspec errors for insert_hoptoad_javascript_notifier.

Re: Wear Deodorant

cobaltchris — Sat, 31 Jul 2010 11:43:31 -0000

So true! We experienced the same in Europe. However, there was one point in Venice, that well nearly sent my wife over the edge... We were on a water taxi, and it was fairly busy. It was summer, short sleeves, etc. My wife brushed by some dude, and literally, he left a sweat trail on her. She was completely disgusted (not unsurprisingly!). Dude was not exercising or anything, just looked like your average guy taking the taxi. Pretty disgusting. Apparently this was the super-stink boat, because we then sat down, next to two cute young ladies (20-ish). We almost had to move they stank so bad! Ladies - no matter how beautiful you are, if you stink that bad...

Re: The Basho Blog: Schema Design in Riak - Relationships

cobaltchris — Wed, 07 Apr 2010 10:52:16 -0000

Thanks Sean. I signed up for riak-users yesterday and had started reading posts I found on number of buckets vs. number of keys and so on. At the moment my app is implemented with MongoDB and is working fairly well, but Riak has just felt more natural in a lot of ways to me, even though really I could use the same style of data model in Mongo or Riak for my app, etc. Partly this is an experiment for me to help get more real world experience with NoSQL db's and so on. I've also found that the Twitter data, while seemingly simple at first, is really quite a fun puzzle at times in terms of how to store it. Things can get big fairly quickly when following social graphs.

One very appealing aspect of Riak is what appears to me to be rather trivial ability to scale up (unlike many of the others). As someone who tends to work either as the only developer or a developer on a very small team, and in keeping with staying lean and low cost, it's quite appealing to be able to simply add another node and not have to manually re-balance things, or shard it simply for the sake of machine size, and so on. I also find that Riak, for my brain, works sort of like Redis on steroids (I originally wrote this app, when it was just for me, using Redis, and really dug that for the simplicity, but it broke down when going to multi-user).

Anyway, I'm having fun exploring DB's other than the usual MySQL and such, and really enjoying working through modeling various app's data in these different ways.

Re: The Basho Blog: Schema Design in Riak - Relationships

cobaltchris — Tue, 06 Apr 2010 14:56:44 -0000

Thanks Sean. Is there a better venue for discussing this data modeling? I'd love to get into it more. As for the statement about "thousands or more" is that dangerous territory just for links I assume? or, I'm a bit confused on that vs. avoiding running map-reduce across a bucket, etc. Maybe in relation, is there a danger level for number of buckets? e.g. if you don't store all known tweets (or similar kinds of data) in one bucket, should you have a tweet-bucket per user?

I realize tons of this is app-specific, but I'm trying to get some guidelines. I can see how to model my app's data in a few different ways, and the blog post helped refine that, but I'm still a bit fuzzy on cases where you have thousands (or more) of things. It sounds like, if I'm not needing to query across those, then storing them all in a bucket is fine. But then doing relationships/links, according to the above blog post, it sounded like I didn't need to worry about say thousands (single digit, or even tens of thousands) of links, but that once you get to say six digits, that's going to fail due to limits on the headers. I do realize that probably having say 10,000 links is probably not a great idea either, but in general, so I'm curious about how you'd structure a relationship say between a User and say the tens of thousands of Tweets or Widgets or whatever that they "own". I could see UserTweets bucket, but then if you get into say hundreds of thousands of users, is that too many buckets? I may for no good reason be thinking that buckets are for very high level concepts, but in reality maybe having many many buckets is just fine?

Re: The Basho Blog: Schema Design in Riak - Relationships

cobaltchris — Sun, 04 Apr 2010 19:56:56 -0000

Thanks for this post - this was super helpful in helping me think about how to model data using Riak, and the whole notion of using different buckets, with the same key (e.g. in the Twitter data), is cool, and useful, and I think fits what I'm after nicely (enough that I'm now thinking of switching this fledging app I'm working on to use Riak :)

I'd like to see you continue with the bit about tweets in particular - or anything that a user would have say thousands or more of. With the above, it made me wonder, well, would you model tweets by having a "statuses" bucket, that holds all tweets your system knows about, with a key that was essentially irrelevant (but likely the ID of the status from Twitter), and then have a "tweets" bucket, where the objects in it were much like the "following" bucket - key of the user's name, and then that object is simply a set of links to items in the statuses bucket. That way you can pull the user object without their tweets, but easily then access their tweets by pulling their object from the tweets bucket. Or, would it be better to simply have say a MapReduce that iterated the statuses bucket and found all the tweets belonging to a given user?

Re: Hopping in the cloud

cobaltchris — Wed, 31 Mar 2010 17:42:50 -0000

Thanks for a great writeup, and for Hoptoad. We are going to be making this same switch in the near future, so your writeup was very timely and useful to us (we're on EY private cloud right now, and have some similar app issues (heavy email, etc.)). Two things I'd love to hear more details on are:

1) specific Exim/SendGrid setup. We already use SendGrid, but send directly to them. But, given we're wanting to send tens of thousands of emails within a couple hours, that doesn't work (on a single slice), and the Exim local MTA idea has come up. So, if you could talk more about that setup, and performance info (e.g. at what rate (e.g. emails per minute) can you push to Exim through ActionMailer (if use AM), that'd be great.
2) did you use Engine Yard's services to help with all this, or did you guys do it all yourself?

And, if it's possible to share, I'd be curious to know what slices you had prior and what you wound up with on the cloud (e.g. how much RAM per slice, how many slices, vs. what EC2 instances you're now using). EY said they'd help us determine this, and we're also planning to do forked traffic to do real world testing, but having more data ahead of time I hope will help us avoid any issues in transition.

Re: Cucumber directory convention

cobaltchris — Sun, 16 Aug 2009 22:07:31 -0000

I get that at a high level, but I'm curious - do you then write what is a similar story for all of your different users? Are the step definitions (where possible) shared? There are a lot of cases in our suite where I've typically mixed users, because the story seems to be written nicely where I can say "and regular user sees...", then "and an editor sees...", then "and an admin sees...", where essentially you're just re-logging in as a diff user. I can see the niceness of the separation on one hand, but I can also see that now you have 4, maybe 5, or more (or however many user types your system has) versions of the same feature story to maintain. Again, not saying it's wrong, just discussing out loud if you will...

Re: New 29er Tires from Bontrager

cobaltchris — Mon, 27 Jul 2009 12:23:57 -0000

Love2Ride, I'd go with the 29-3. I ride Mtn King 2.4's in the winter - but they don't have the volume of the 29-3's and I feel the 29-3's roll faster, and grip as well (better in some conditions, not as good in mud or slop).

For a rear, I haven't ridden the 29-3 rear tire, but people seem to like it quite a bit. It looks totally different than the front, but is in a size class similar to others you mention, and apparently works really well. I wouldn't touch Conti Race King's, simply because they're too small for what I want (afterall, I was favoring the FR-3's or 29-3 (front tire) front and rear on my rigid 29er :) My take is that unless you're racing, I'd be running something bigger as it's your suspension, better traction, better braking, better cornering, uh, well, you get the idea :) But if you favor a smaller rear tire, then try the 29-3 rear. I haven't ridden the XDX. The Nanoraptors are ok, but weren't enough for me.

Re: New 29er Tires from Bontrager

cobaltchris — Thu, 16 Jul 2009 12:10:43 -0000

I would take the 29-3's (front AND rear) over the WW's because they mount up tubeless with utter ease, and that lets me run them pretty low pressure. The size is pretty close to the WW 2.55's, close enough that the comfort aspect is still there, and I think the tread pattern works better as well - gives you more range. I also would think the 29-3's would be a lot faster over the terrain you're talking about, they roll exceedingly well.

If I lived in Bend, OR, which is nearby to hear, but has conditions like you mention, the 29-3 would be what I would run, and I mean I'd run the 29-3 front tire both front and rear.

Re: Does Anyone Use Chrome?

cobaltchris — Thu, 09 Jul 2009 00:11:06 -0000

As some have said, the browser percentages will obviously depend a fair bit on the nature of the site. I suspect you have a higher than average number of digerati/tech-savvy folks visiting your blog (and with Firefox being your top browser, that proves that IMHO). Sadly, IE still wins on very broadly appealing sites. At DealBase (hotel deals), we have a higher level of traffic, but it's over a very wide range of users. Chrome sits in 4th for us as well, but only at 2%.

Granted the Fake Steve article is of course a lot of fun, but funnily enough Chrome is of course related to Safari, so that's a bit funny. My hope is that Chrome actually does help erode IE. I don't use it more than to test our site with it, but I would like to think that with Google's clout, that maybe it can help open more people's eyes to something other than IE (for the folks that would otherwise essentially not pay attention to, or not really understand that there are other browsers). We'll see. Google isn't yet promoting it, but maybe once it's out of beta (hmm, hopefully not as long as Gmail ;-) they will...

Re: SelectorGadget Bookmarklet

cobaltchris — Fri, 13 Mar 2009 16:29:08 -0000

Superb, thanks so much for this, will be very useful!

Re: ReinH — A Git Workflow for Agile Teams

cobaltchris — Tue, 03 Mar 2009 21:28:09 -0000

Rein, good writeup, basically the identical workflow we use at DealBase.com. We too use Tracker, and I require the tracker story in commit messages, etc. With that, I wrote a little GitHub-Tracker post-receive hook service, so that your commits that have tracker story ID's in them will add your Git commit message as a comment to the Tracker story. You can also update the Tracker story's status with this as well. You can get it or fork it here: http://github.com/chris/tra...

The format is fairly similar, but instead of "#" I use "Story" because we had other things that were using post receive hooks and I needed it to be unique. So, the service just requires you put something like the following in your commit message:

[Story234567]

[Story234567 state:finished]

The nice thing is of course that this won't get picked up by the service until you've pushed to the upstream master (GitHub), and so, I think works well in that it won't mark your Tracker story finished until your code is truly in GitHub and available to everyone. This works for our workflow anyway, in that we mark it finished once it's in GitHub, and then it gets marked Delivered in Tracker once it's been deployed to our test server (after it's gone through the continuous integration server and been green lighted).

Re: Lua, the Embeddable Game Scripting Language

cobaltchris — Wed, 28 Jan 2009 01:04:25 -0000

Lua is interesting, and great for embedding as you said. Just as an interesting note about it, Adobe's Lightroom is written partially in Lua, and extensions for it can be written in Lua (or could the last time I checked, but I haven't worked at Adobe for a while, so not sure about Lightroom 2.x).

Re: Fluid App Icons for Mint and Pivotal Tracker

cobaltchris — Tue, 16 Dec 2008 16:23:44 -0000

Thanks for the icon. Did you do anything special to get Tracker to work in Fluid? I created a Fluid app, but anytime I try to login in the Fluid app, it opens Tracker in my regular Safari instance instead.

Re: RSpec should each matcher — Thoughtfolder

cobaltchris — Fri, 05 Dec 2008 01:07:10 -0000

This appears to be wanting Cucumber to be installed, maybe via one of the gem dependencies? Ideas?