Relentless Simplicity :: The Bonanzle Tech Blog

I did not find anything to love about the results I was getting searching for queries like “Increase rubymine memory” and “set rubymine memory”. Here’s the location of the file that dictates Rubymine memory usage in my Ubuntu install:

[RUBYMINE_DIRECTORY]/bin/rubymine.vmoptions

Inside you can specify lines like:

-Xms800m
-Xmx1200m
-XX:MaxPermSize=1000m
-ea

These declare maximum memory usage, maximum sustained memory usage, and, uh, some other stuff.

[Hops onto soapbox]
I love Rubymine, but I sometimes wish that instead of adding a ridiculous number of new features, they’d just make the damn thing run consistently fast, and not make me worry about memory usage. I know that with my web site I always have a preference toward adding cool new stuff, but responsible thing to do is often to perfect the pieces that are already there. Almost everything negative I’ve heard about Rubymine (which isn’t all that much) is along this same vein, but I’ve never heard the Rubymine team admit that it wishes that the product ran faster or its memory management were better.
[Hops back down]

For my money, still definitely the best choice for Linux IDEs.

Bill Uncategorized Rails

Or do you notice that every time you visit a website or a submit a form, and you get the indefinite spinner of doom, that the url always seems to end in .aspx?

Hmmmm!

Bill Uncategorized

We want it bad. Reeeeeal bad.

Bonanzle is still running on Subversion (via Tortoise), because comparing its UI to the UI for the de facto git standard (msysgit) is like comparing Rails to ASP on Visual Basic. Yes, the difference is that big. Msysgit is ugly as Pauly Shore, most of its windows can’t be resized, it crashes regularly, and trying to decipher the intent of its UI is like reading dead sea scrolls.

Yes, yes, I know: if you don’t like a piece of open source software, you should shut up and fix it. Unfortunately, I’m sort of single-handedly maintaining this web site that has grown to about 150k uniques this month, where two months ago we had about 10k uniques. I can not end world hunger and rescue my cat stuck in a tree.

But if there is any intelligent life out there that has spare programming cycles and the desire to make a huge difference in the world, this is my personal plea that they will consider giving some love to the woebegone msysgit… or maybe just start their own Windows git client, I can’t imagine a real hacker would take more than a week or two to match the featureset of the existing msysgit.

I’d really like to move Savage Beast to github, and I’d really like to collaborate on the other projects that are happening on there, but it just doesn’t make sense to go from a slick, error free, decipherable UI like Tortoise to the meager helpings of msysgit.

I’d happily donate to a project like better Windows git.

Preemptive note to smart alecks: No, I’m not moving to Mac (or Linux) now. There are plenty of reasons why, I’ll tell you all about it some other time. Incidentally, what is the preeminent GUI for git on Mac these days? From what I understand, many of the real hackers are perfectly content using git from the command line…? Shudder to think of reading the diffs and histories.

Bill Uncategorized

I’ve had a devil of a time trying to get Google to tell me how to write a Mysql query that allows us to somehow perform a MySql query that 1) filters rows on a distinct column 2) returns other columns in the query besides the distinct column and 3) allows us to order by a column.  In our case, we (and you, if you’re running Savage Beast!) have a list of most recent forum posts on the site — currently, if you list all recent posts, the search will just find all posts and order by date of creation, but this makes for some dumb-looking output since you often end up with a list where 10 of the 20 posts are all from the same forum topic.  All the user really wants to know is what topics have a new post in them, and to get a brief glimpse as to what that new post might be.

Thus, we want to create a query that returns the new posts, ordered by date of creation, that have a distinct topic_id.

Here’s the SQL that can make it happen:

Post.find_by_sql(”select posts.* from posts LEFT JOIN posts t2 on posts.topic_id=t2.topic_id AND posts.created_at < t2.created_at WHERE t2.topic_id IS NULL ORDER BY posts.created_at DESC”)

Hope that Google sees fit to lead other people here instead of struggling to get GROUP BY to order results beforehand (GROUP BY ‘posts.topic_id’ works, but it returns the first post in each distinct topic, rather than the last post as we desire), or get SELECT DISTINCT to return more than one column, as many forum posters unhelpfully suggested in all the results I was getting.

Update 11/26/08 - A Word of Caution

I finally got around to setting up some profiling for our site yesterday and was surprised to discover that the above query was taking longer per execution than almost anything else on our entire site.  The SQL Explain was not too helpful to explain why, but it showed three joins, the join on the topics table involving every row of the table (which is presently almost 10,000).

Takeaway:  for this query to work, it seems to consider every distinct topic in the table, rather than being smart and stopping when it hits the per-page paginated limit.  Since I already determined that “group” and “distinct” were non-starters for being able to pick the newest post in each topic, I ended up revising the way the logic was done to an easier to manage and far-more-DB-efficient way:

We now track in each topic the newest post_id within that topic.  While this adds a bit of overhead to keeping the topic updated when new posts are made to it, it allows us to do a far simpler query where we just select the more recent topics, joining to the most recent post in each topic, and then ordered by the age of those posts.

If you have the ability to create an analogous situation to solve your problem, your database will thank you for it.  The above query starts getting extremely slow with more than a few thousand rows.  Yet, I defy you to find an alternative to it that works at all using “group” or “distinct.”

Bill Uncategorized

With so many sites trying to be “the next Craigslist” these days, I thought it prudent to take explore what’s going through their minds, and why it is that they keep on failing.

The answer why lies in my Seattle Craigslist vs. Microsoft Expo, or, Don’t Ape the Mojo blog on the mothership.

Bill Uncategorized

To add a user group, see http://www.cyberciti.biz/faq/howto-linux-add-user-to-group/

To view a user’s groups, use “id [user]”

Hard to Google.

Bill Uncategorized

http://www.bonanzle.com

There is no way you could post items for sale faster anywhere on the Internet.

Please support this blog and put some stuff up for sale. The time to take pictures + visit the site + post the items is nearly guaranteed to be less than an hour in total.

Bill Uncategorized

By far the best book I’ve gotten to help with all the not-well-documented aspects of getting a site delpoyed. Beginner Rails help articles are a dime a dozen, but this book is chalk full of example-laden recipes for deployment. You are a fool if you deploy Rails apps and you don’t own this book.

Bill Rails, Uncategorized

This comes from an anonymous programmer applicant, who had the following assessment of my last posting about working at a startup:

“I had in fact already read your blog on startup environment. I have not experienced a work setting of precisely this nature, except possibly as an EMT on scene. “

Bill Uncategorized

You can look it up on www.williambharding.com — there are few things that I love more than seeing a plan hatch, and that’s just what I’m seeing as this site makes some big steps toward being the most usable site of its kind for its users. We began our first round of seeking seed investment this week, and so far, so great. People seem to quickly comprehend how this will be a big step forward in shopping, and the fact that I have a team that personifies awesome certainly makes for an easier sale.

Once we finish acquiring our seed funding, things get really fun. The old adage of “time is money” inverts to state “money is time,” and my hope with seeking our first group of investors is that the money we can obtain will buy us many months of time compared to the means that I’ve used to get the site to where it is now. I’ve been fortunate to get significant discounts on basically everything we’ve purchased so far (web design about half price, programming costs ridiculously low for the quality) by getting people excited to be part of the idea, but goodness, having a bit of capital to work with is going to open a lot of exciting new doors.

There’s still a hell of a lot to do, but if the team matters as much as the product, we’re golden.

Bill Uncategorized