• Button-click handlers are now executed without blocking the rest of Resolver One. This means that if you accidentally write a handler that never finishes, or just one that takes longer than you want, you can cancel it while it’s running and fix the problem.
• We’ve added a dialog for setting wrapping and aligment for cells.
• External imported modules now shared between documents and RunWorkbook (which gives much better performance for RunWorkbook).

Here’s the full list of changes — and, of course, you can download it from here.

This spreadsheet uses the most simple parts of the [TheyWorkForYou] API to download a list of every constituency in the country, and then help you get some basic information about each, including a link to a Google Map showing where it is.

More here.

NodeXL is an Excel add-on for drawing network graphs. It is becoming extremely popular and I was wondering if anybody has an implementation for Resolver One (is it possible at all)

I have written a script for connecting Resolver One and GraphViz but NodeXL looks really slick

NodeXL

I have developed a little Resolver One spreadsheet that can update your tweeter account, get your friend’s list or get the list of those who are following you.

It’s fun please try it

Download it from here

For my masters I wrote a 3D visualizer in Macromedia Director for the defense department to visualize ammunition bunkers. I always wanted to export my Master’s Thesis to a better 3D environment, but never found any time to do so

I saw people were doing OpenGL in RO and since I don’t know any openGL I started writing a DirectX code to do the inventory visualization. BUT I CAN’T

has anybody done any DirectX in resolver one? please help me

Here’s the full list of changes — and, of course, you can download it from here.

=SUM(quantity * price for person, quantity, price in zip(#Who#_, #How many#_, #How much#_) if person == 'Fred')

For more information, see:

• Our documentation for header rows.
• Python.org documentation about list comprehensions.
• Python.org documentation on the zip function.

There is a rumor about the origin of Resolver One. Some people say that the main reason RO founders came up with the idea of resolver one was to be able to play super mario on their windows machines. That’s basically why Resolver Systems is investing so much in IronClad because, for running a good mario game you need NumPy. Anyways I don’t know whether or not that rumor is true. But I know that there is a version of super mario for Resolver One and I got my hands on it yesterday. Here is how you can run super mario on your Resolver One. It is a little slow though :)

• Download this jar file iMario.jar
• Go to the directory where you downloaded the Jar file and type the following to run the TCP server
  java -jar iMario.jar -server on -vaot on -vlx 1044
• Download this file and run the rsl file in the root directory

For more information on how to write your own client see this page

Over the course of 8 months I have been refining my Yahoo Pipe for Resolver One. Yahoo pipe is a service that can help you combine news sources, add filters and export them as a live RSS feed that people can add to their Google Readers.

This Pipe is more efficient than Google Alerts. For example Google Alerts do not show you conversations on twitter about Resolver One but our pipe does. It has been a reliable source of information for me to get up to date news about Resolver One. I have blogged about it before but wanted to mention it again. You can modify this pipe too and if you got a better pipe after modification let me know so I can use it myself.

If you want to add it to your Google Reader (or any other RSS reader) just use this link.

I was working on a lot sizing problem today and implemented a Wagner-Whitin algorithm in Resolver One. I just want to share it with you here. Some people use a MIP solver to solve lot sizing problems while Wagner and Whitin have devised an algorithm that can be implemented quickly on a spreadsheet and there is no need to use an expensive solver. You can download the spreadsheet from this link.

In the classical lot sizing problem we have the following assumptions:

1. We have a single product (with the production quantity in period t)
2. Discrete Time (unlike EOQ model)
3. Demand is known but varies over time (e.g., we have orders)
4. No shortage is allowed
5. No capacity constraints
6. There is a positive setup cost that might be varying over time
7. The unit production cost is constant
8. The objective is to minimize the total setup cost and inventory costs

Here is the formulation for the model:

1. setup cost in period t
2. inventory holding cost per unit remaining at end of period t
3. demand in period t
4. a large number (will be used in the MIP model)
5. production quantity in period t
6. if and is 0 otherwise
7. inventory remaining at the end of period t

The MIP model will be:

Wagner and Whitin have solved this problem using the forward dynamic programming. In this spreadsheet we implement the forward dynamic programming for the lot sizing problem. The spreadsheet then suggests the time period than it is optimal to pay for the setup cost, in other time periods it is cheaper to carry an inventory.

In our institute we will be holding a symposium about the future of design (we will have fascinating speakers like the chief designer of IDEO and Smart Design. send me an email if you are interested, it is free for academics and pretty cheap for others, all the talks will be open source and we will post them online afterward)

I volunteered for preparing the name badges. To be honest I didn’t really realize what I was doing when I volunteered. I got a messy excel file with 500 records in it and I needed to design and prepare 500 name badges. So I am going back to two of my old friends to help me with the this name badge mess (GDI+ and Resolver One). I will upload my file either here or on the exchange when it is ready. I am slowly noticing the value of having a flexible tool like resolver one for quick tasks like printing these stupid name badges.

I will report my progress here. and will post my file later (if I can finish it).

The big news in version 1.6 is that we’ve made it easier to share your spreadsheets with people who don’t have Resolver One:

• Player licenses. These new free license keys allow people to use Resolver One in a mode where they can use the spreadsheet functionality, but can’t see or change the Python user code. So you can build spreadsheets and pass them on to other people, safe in the knowledge that they won’t have to buy Resolver One in order to use your work. You can also switch your own non-Player version into a player mode in order to see how it will look for other people.
• Resolverlib: library support for executing Resolver One spreadsheets from IronPython scripts, so back-end processing scripts built by IT teams can use logic designed in a spreadsheet model. More information on Resolverlib here.

And as always, there are many more useful features, including:

• Simple auto-complete in the Console — try typing workbook["Sheet1"]. (including the dot)
• NumPy support is out of beta! In Resolver One 1.6, if you add import numpy to the start of your pre-constants user code, NumPy will just work — no need to mess around with environment variables.
• Expanded cells. Make a cell grow to cover other cells below it and to its right, giving you much more control over your formatting. A much-requested feature that really comes into its own when combined with…
• …images in cells! If you set the .Value of a cell to a .NET Image, then the image will be displayed in the cell. This is useful on its own, but when combined with expanded cells, your spreadsheets can suddenly start looking much more attractive. See both of these in action in our “Histograms in cells” example.
• To make creating graphs easier, we’ve also added the popular ZedGraph package to Resolver One.
• And, of course, a number of bugfixes and performance enhancements.

If you want more details, here is a full list of every change in this release. Don’t forget, we have forums to discuss and ask questions about Resolver One, and there is documentation on the website too.

Ironpython in Action

I think it is safe to assume that all of the readers of this blog know IronPython. I have been skimming through random chapters from “IronPython in Action” by Michael Foord. Here are some of the chapters that I really liked and look interesting and I was wondering if other people here have a favorite chapter. Feel free to share your favorite chapter too.(I actually want to know which one I should read more carefully ;)

Perhaps in homage to its roots in finance, one of the entries to the Resolver One Spreadsheet Challenge, and the third spreadsheet we are looking at in this series of blog entries, is a spreadsheet application for Stock Trading and Portfolio Analysis. It is the creation of the winner of one of the earlier competition rounds, Siamak Faridani.

• Stock Trading and Portfolio Analysis spreadsheet

Resolver One is a “mash-up” of a spreadsheet and an IDE -Wikipedia

Resolver One is supposed to be more powerful than a spreadsheet and more flexible than an IDE. In this entry we demonstrate how a sophisticated trading software can be developed in Resolver One quickly.

It can generate candlesticks, finance bars, and line plot. It simulates the value at risk with two methods. And can be connected to R for portfolio optimization.

There are many impressive things about this spreadsheet, not least of which is the PDF documentation that comes with it. The main dashboard, part of which is shown below, allows you to create a portfolio consisting of positions in several stocks, and then perform a range of calculations or present information about the portfolio. Calculations include the rate of return and Value at Risk analysis.

Once you have created your portfolio historical price data is fetched for all of the stocks from google finance. If you have the Resolver One Financial Data Feeds enabled then it would be easy, and faster, to use Bloomberg or Thomson to get this historical data. Price data is put into new worksheets, created and formatted from code, for each stock. The one improvement I would make is to cache these results, using a Cached Worksheet, to avoid having to refetch them every time the spreadsheet is recalculated.

The spreadsheet is capable of generating a variety of different charts for the portfolios it manages. The image below is a pie chart representing the whole portfolio.

This is a data plot (a candlestick visualization) for one of the stocks; Google Incorporated in this particular case.

The spreadsheet has several other nice features, including the ability to fetch recent financial news relevant to the portfolio or take you to the google finance and bloomberg pages for individual stocks. This is definitely fun to play with and a good example of some of the things you can achieve when you combine the spreadsheet interface with complex user coded functionality.

The first podcast is with FLOSS weekly presented by Leo Laporte and Randal Schwartz.

• FLOSS Weekly: Episode 82
• Direct link to the mp3 download

The second podcast was recorded a few weeks ago at the EuroPython 2009 conference, which Resolver Systems was pleased to be able to sponsor. The Linux Outlaws recorded a series of interviews at EuroPython, including one with Michael Foord. The interview with Michael starts fifty five minutes into the podcast. In this interview we put out a call for developers to help us port Ironclad to run on Linux with Mono:

• Linux Outlaws 106: EuroPython Special
• Direct link to the mp3 download

The second top keyword on google that people use to get to this blog is “resolver one MATLAB”. Thus, I thought it would be good to share my experience with MATLAB and Resolver One. So far I haven’t done any real application with MATLAB and Resolver One together but I have experimented a couple of ways to integrate these two.

The best way is to use “MATLAB builder for .NET”. Builder can compile your MATLAB as a .NET library. There is one catch though, builder does not come with MATLAB. Although you can distribute the compiled .NET library freely to computers that don’t have MATLAB. MATLAB builder can be tricky but it has a very good documentation that walks you trough building .NET libraries.

When you don’t need the real time integration one trick that I have used is to write .rsl files directly from matlab. rsl files are textfiles and have a very easy to understand structure [as far as I know, there is no official rsl specs anywhere on the net] This method is very convenient and works perfectly in many case.

The third method is to use Octave instead of MATLAB (they share the same syntax) and then use the R example on resolver exchange to connect it to Resolver One.

PS.

Here are some of the keyword that people have used to get to this blog (I filtered out irrelevant keywords). The top keywords are “Poisson point processes/and simulation” the others are “resolver book” and “resolver design pattern”, “resolver one tutorial”, “resolver one simulation”, “Ironpython excel” “resolver one portfolio optimization” and “resolver one matlab”

2. A dear friend of mine has got me a copy of “Ironpython in Action” my plan is to read a couple of chapters this weekend (I will be on a trip for a couple of days). It seems that unlike many python books that have only a number of pages about .net, this book has a good blend of material about Python-style coding and .Net framework stuff. I am still wondering why this book is not part of the safaribooks online while many of manning’s titles are there (we have a subscription of O’reilly’s safari books at our institution).

Will wright is certainly one of the master minds and one of the greatest creators in the game industry. We went to his lab today to see some of his latest creatures. Interestingly enough he is developing robots now.

He talked about mind and how (he believes) the brain works. I liked the man. Very geeky indeed but he showed that he has earned his reputation for a reason. He seemed to have a great understanding of where we are heading to in terms of the software industry. I had nothing today except so much respect for this great man.

He also talked about the object oriented model in Sims and why he calls it “subject oriented”. SIMS has a great OOP model I will try to find a link that explains it and will post it here.

All in all I have realized that when people run a multimillion dollar company and don’t get to worry about paying bills they get to have more fun and will get really creative.

Update: His conversation with Colbert

1- David Karger is a MIT professor and one of the greatest minds in the business. Now see on which weblog he has made a comment :) I told you we will succeed. we now just need to get Richard Karp to write a comment too!

2- I have been working a lot with GNUplot lately. I just love how it integrates with the rest of your code and its templates are fantastic and flexible. I found this book very useful if you work with GNUplot (“GNUplot in Action from Manning“) you don’t really need to buy it, just take a look at a couple of first pages online and then go to GNUplot website and play with examples (alright if you have the money go ahead and buy it, it is a good book, the guy has a PhD in theoretical physics)

3- Michael Foord has looked at one of the eateries of the Exchange. It is a great entry take a look at it. if Michael continues to do that I can stop introducing entries on this blog

4- I am writing a paper for CHI2010, and am using Resolver One for analyzing the data. I will post a draft here before the deadline (Sep 17th)

As you know you cannot import huge dump files to your MySQL via phpMyAdmin’s webbased interface. If you are working with huge datasets and want to import your datasets into your mySQL engine big dump is for you.

It helps you import huge data sets via web without going through nasty SSH commands.

Correction: Half way through my 1.6 GB dumpfile big dump broke and quited with an error massage, I am back to the good old days. Crappy SSH command it is.

I am conducting a usability study for the application that we have written for the state department. As a courtesy to our participants I give them $10 amazon gift cards for one hour of intense UI study! today I read on Haystack blog that they are giving $30 to participants simply to read blogs via Google reader.

I guess I need to buy more expensive cards otherwise I feel guilty of abusing our participants :(

here is a simple example of how you can load images into your cells in the new Resolver One

Download the file from here

Resolver One 1.6 with a number of nifty features is ready for beta testing, I was talking to another RO user yesterday and she was more excited about the “images in cells” capability. She was always complaining that her spreadsheets look crowded with many graph tabs.

I am personally excited about the “player” version. Since it shows that RO will remain a great tool for developing real systems for different clients.

Download the new version from here http://www.resolversystems.com/beta/Resolver-One-4608.msi

and this is a short list of new features:

• Expanded cells
• Images in cells
• Simple auto-complete in the Console
• Player mode
• NumPy support
• Fonts
• Drag to re-order worksheet tabs

Like any other beta version it still has some bugs. but who cares Windows still ships with bugs :)

I am working on a post about how to customize graphs in Resolver One. It is basically an explanation of how these entries work (+,+,+,+, and even 3D ones +). The post will go online later this week.

Meanwhile, related to the previous post, a reader asked me how he can get better in Flex. Here is the book I have used and recommend. It does not make you the greatest Flex programmer but you’ll be pretty good after working with it for a couple of months. “Action script 3.0 Cookbook” by Joey Lott

ActionScript 3.0 Cookbook: Solutions for Flash Platform and Flex Application Developers

Flex is adobe’s most powerful programming platform for developing Flash-based desktop and web apps; and R is hands down the most powerful statistics bazooka. I have been working with two of the Resolver Exchange entries (Resolver One/R, and Resolver One/Flex) the former enables you to connect Resolver One to R and the latter helps you connect Resolver One and Flex.

As far as I know there is no library for R that directly enables you to develop Flash content from your data (you typically need to use RApache to do so) By combining the above libraries together I could get my R codes to generate eye catching interactive Flash graphics. Underneath, I am using Resolver One as both a database engine and a middle ware to keep both elements in place

I am preparing a demo for the upcoming UseR meet up of the Bay Area R users group. I will post my samples and powerpoints on Resolver Exchange soon.

Note: Getting Resolver One to work with Flex Builder is a little cumbersome, if you had any problem installing it please shoot me an email and I can help you. Furthermore, if you are used to SPSS there is an R add-on released just recently that let’s you use R the way you are using SPSS (see Andrew Gelman’s post)

Note II: One problem that I always have is that Flex compiler is very slow it takes 30 seconds for my machine to compile a relatively large Flex project (~15,000 lines of codes)

My friend George Markovitz died a year ago, he was a true UNIX geek. He started writing codes on punch cards and then moved to Vax and IRIX systems. His favorite field was computational fluids mechanics and he used to lock our systems up for days to run his simulations (I used to hate him those days, he was really good in what he was doing)

When you are running simulations for days on a remote computer the most important thing is to know when the process is finished; otherwise you have to check every hour to see if your remote computer is still doing the number crunching.

Last year he gave me a piece of code that has helped me more than any other code in my career. And I want to share it with you. This code will text message/email you when your process is finished. Here it is:

PID=`ps -eo pcpu,pid -o comm= | sort -k1 -n -r | head -1 | awk '{print $2 }'`
while [ -e /proc/$PID ] ; do
sleep 600 # chage it to 30 or any other time in seconds (600 seconds=10 minutes!)
done
ifconfig | awk '{print $2}' | head -2 | tail -1 | sed 's/addr://' | mail -s "Process done" youremail@gmail.com

save it as mailme. and run it with $nohup bash mailme&

in your gmail change the setting so that it sms’s you whenever a new email arrives and you are done
When your machine is finished with your task it will send an email to your gmail and gmail will sms you on your cellphone. This way you have a great infrastructure for your number crunching needs.

To run this on an Ubuntu linux you need to have mailutils and postfix; simply install them with

 sudo agp-get install mailutils

This line sends emails

echo “Process finished” | mail -s “process finished” youremail@gmail.com

I hope you enjoy it.

George you’ll be missed!

Sorry about being scarce these days. I had to drive to Michigan and was exhausted when I got back. but here are some community news: A thoughtful article about spreadsheets and Resolver One has been published on Resolver One news. And the winner of the last round is announced (Don’t forget that Sudoku in general is NP-complete but some instances of Sudoku can be formulated as network flow problems).

Our friends in London have decided to give some love back to the Resolver One community by posting all of the entries for the final round of the resolver competition online. Entries are all interesting and they can be all potential nominees for the final winner (Although I am still rooting for the Resolver-R integration since I have used that one a lot)

Update: those of us who live around bay area/San Fransisco are thinking about having a Resolver One meet up soon, we will have a very informal meeting on either one of the Stanford or Berkeley campuses; And we will share ideas and experiences. Our group is very small right now but we are hoping to get more people, especially we are looking for those who use it in their financial firms. Feel free to drop me an email if you like to join us (we will have free refreshments)

I know there is a spreadsheet on resolver exchange that connects Resolver one and R. I have been battling with that spreadsheet to understand how I can connect R and Reinteract.

I basically want to have something like reinteract for R, any ideas?

I am very excited about the upcoming release of Resolver One. They report that they will have:

1. In cell images
2. Resolver Player

Resolver Player supposedly is a resolver one without the code view section (like Mathematica player) it is great specially if it is low cost. I also like to see a closed source binary rsl as I may not want to share the source with the client.

The “in cell images” is also very exciting. Looking forward to working with it.

Send them an email if you like to try these new features beta.signup@resolversystems.com

Those who think that Python and IronPython are exactly the same have done neither GUI programming in Python nor error handling in IronPython – Abraham Lincoln

For large datasets the only way you can process them in R is to load your dataset in a MySQL database and then connect with R to MySQL to process the data. I am now processing a dataset with 3 million records and I imagine it will take 3 hours to process (I am doing a hierarchical clustering on subsets of this dataset)

One question that I always have is “how can you process terabyte size datasets in Resolver One?”

As you know Google has recently released it’s very own LP solver under Apache platform, they have been using it on their online spreadsheet application. I was wondering if anyone is up for integrating it with Resolver One?

Email me if you are interested or go ahead and code it up and send me the link to your code

It would be a great product

Update: The Google SiplexSolver only uses the two-phase simplex which is very inefficient (as far as I recall the revised simplex is better and has the complexity of O(mlogn). For a huge system a combination of interior point and simplex is always better.

“One month of software development consists of one day of programming and 29 days of refactoring” – J.D. Salinger

I just submitted my last entry for the Resolver One competition. Regardless of any future outcome I am extremely pleased with my entry and the whole competition and I hope my work comes handy to you too. I have been working on it for almost two months and I am sure many other people have spent so much time on their rsl codes and I am eager to see other codes as well. I have learned a lot from Resolver folks and I hope we can keep hacking with Resolver and introduce it more to our colleagues and friends. After all we all love this product.

I have been very sick for the last couple of days and I guess I need to sleep for at least 7 days. Wake me up whenever the winners are announced :) I will publish some more Resolver One tiny codes later this week.

Good luck to anyone who has participated in this 6 months competition.

PS. By the way, my favorite economist, Kevin Murphy was interviewed on this website. I’m honored that I have grown to call him my mentor and a great friend.

Ah, Haskell.,

Alas people don’t know you

- William Shakespeare

For a finance person no day can pass by without touching R. In this post I’d like to point out a number of helpful R commands. For running R in Resolver One please use this entry.

In this tutorial I work with a csv file (I want to rearrange the columns and plot a histogram of eigenvalues)

Reading a csv file into R:

to read user selected a csv file from a file dialog use

• usertable = read.csv(file.choose())

read.csv automatically considers the header. To see a couple of first rows in your variable use the following

• head(usertable)

To see a list of variables in the memory type “ls()”. If you have a MATLAB data file you can also use it in R by typing “readMat()” note that you need to install R.matlab package. To clear the memory before running your R script you can use “rm(list = ls())” this is eqaul to “Clear all” in MATLAB.

Unlike resolver one and similar to matlab an element in the matrix is called by [Row, Col] address. For example the element in the 4th row and the second column of the usertable variable is called by “usertable[4,2]“

For a beep in R use “alarm()” at the end of your code to let you know it is done. The following is a some clustering method ran on my data in R

Data Clusterin R/Resolver One

Neda Soltani

This young lady was killed in Iran yesterday. She was protesting against the fraudulent presidential election with her father when she was shot by a bullet from the riot police in Tehran. If you haven’t heard about the crisis in Iran please consider closing your IDEs and log into bbc or cnn

There is a live blog thread on the huffingtonpost blog as well

Thanks

I have to confess I have tried hard to replace R and Matlab with Python in my everyday life, and I have failed greatly. Just the fact that arrays in Numpy are inconsistent with Matlab is irritating. furthermore, I haven’t found any tool for image processing better than Matlab, and R is too sweet to be replaceable. Another thing that I don’t like about doing scientific computing in Python is that python started as an amazing scripting code just like Matlab and now it is a power full OOP language. It’s too complex for the dirty number crunching (F# shines here, it is the greatest new language for scientific computing that I have seen) I used to love Fortran, Python is much better than fortran but not as good as Matlab/Octave

Ever since I have found this R library for Resolver one my life has been much simpler, I now run some data processing in Python and in Resolver one. If the algorithm that I want to use is too complex I pass everything to R and then run my commands from Resolver One in R and then read everything back in Resolver One and load it into the grid. Using this procedure I can run a K-means classifier in R with my resolver one data.

I know people who have codes in Python, Fortran and C together. My codes are pretty much combinations of Python, Resolver One, R and Matlab (occasionally Java or C#)and this combination works perfectly for me. I have replaced Matlab GUI with .Net gui now but almost all of my data crunching is done on the R end and all the data parsing and pre/post processing is done in Python and ResolverOne

Do you also have mixed codes? If you do how do you test your system and how do you set up your software tests (I am still trying to figure out how to test a code with mixed languages)