However, understand that this instance has some limitations. At present it contains only a small amount of data taken from USASpending.org, which does NOT contain unit quantity information. One of the main purposes of PriceHistory is to compare per-unit price and also to compare multiple sources/vendors, so this demonstration is weak compared to an instance loaded with proper data.
To log into this instance, please use the credentials:
houstonUser99 RpG0pRDyIrhbcBK82a3hqjvir
(This instance has been created primarily to support the upcoming Houston Hackathon (see below)). Feel free to create your own test portfolios.
The name of this project is changing from PricesPaid to PriceHistory. The PricesPaid project has been forked by the government and the fork is being developed in a closed-source manner, although it will continue to be called both "PricesPaid" and "P3" (for Prices Paid Portal.)
In order to avoid confusion with that project, this project is changing its name to "PriceHistory".
In other news, the PriceHistory project will be one of many teams working at the Houston Open Innovation Hackathon of May 31st-June 1st. In preparation for that hackathon and to ease reuse fo the project by other municipalities, I (Robert L. Read) am slowly changing the name and making other changes that should make this project re-brandable and re-usable by a city such as Houston.
In particular, I hope to prepare a vagrant script to ease the rather arduous installation process right now.
The PriceHistory (P3) project is market research tool to allow search of purchase transactions. It is modularized into 5 github repositories:
To learn how to install the system, please refer PriceHistoryInstall project, which contains a Vagrant install script. That repo is actively under development in preparation of the Houston Hackathon.
The name "PriceHistory" is descriptive: this project shows you prices that have been paid for things. However, the name is applied to many projects, so "P3" is the specific name of this project.
PriceHistoryGUI, in conjunction with its sister project also available at github, PriceHistoryAPI, is a system to allow government buyers to research actual prices paid in order to lower the price the Federal Government pays for goods and services. It was initiated by a team of Presidential Innovation Fellows (Round 2), consisting of Robert L. Read, Martin Ringlein, Gregory Godbout, and Aaron Snow.
It is intended in part to address a mandate of Congress, Title 41 U.S. CODE § 3312 - DATABASE ON PRICE TRENDS OF ITEMS AND SERVICES UNDER FEDERAL CONTRACTS.
The main purpose of this code is for a buyer to perform market research of the specific transactions that are related to what you are buying. Individual transactions can be selected by "drag 'n drop" into portfolios which can then be exported to very simple HTML which can imported into a text editor to document market research.
Within the Federal government, this software uses data which is consider sensitive and not available for public usage. However, the code is in the public domain.
This code could be reused by a State or municipality that had transactional pricing data that it wanted to present in a searchable form. It provides a very simple method (documented in PriceHistoryAPI) for importing CSV files in any format and mapping fields to standard, semi-standard and completely customized fields, all of which are automatically searchable once they have been imported into SOLR.
This repo is a Python/Javascript/CSS front-end that ties together 3 other repos : https://github.com/18F/PriceHistoryAPI and https://github.com/18F/MorrisDataDecorator.git and https://github.com/XGov/P3Auth. Fundamentally API is a a simple Python interface to the SOLR full-text search engine. The MorrisDataDecorator is a system for making assocations between records in SOLR. In this case, it implements a "Portfolio" concept. The PriceHistory system does not use a relational database, and has no notion of user identitiy---all portfolios are shared at present.
I've been told I should try to make this PIP installable, but I am too much of a python neophyte to really understand how to do that, so I have opened it as an issue.
At present, this is a WSGI app, and I have only run it with Apahce, although I'm sure some other webserver would work.
The Directory structure that I use is:
- All for projects (PriceHistoryGUI, PriceHistoryAPI, MorrisDataDecorator, and P3Auth) are parallel beneath a single directory (In my case, PriceHistory.)
- I have to add by hand an empty init.py file to that directory (parallel to the others.)
The PriceHistoryGUI module use all of the other three. PriceHistoryAPI uses P3Auth, but does not rely on any of the others. One could easily use PriceHistoryAPI only as a loading tool to load SOLR with data, and then use SOLR directly, or write your own GUI, for example in Ruby.
This is really a visualization tool for .csv files. However, right now, it is tied specifically to the idea of "Prices Paid" for the government and it is logoed for that. We hope to evenutally separate that out, not only so that the code will become more valuable to the open-source community, but also because even within government we have need of separating out the specific style and fields.
The code at present calls the PriceHistoryAPI to search a SOLR database with what we hope will be a "google like" search. The results are then presented to the user to be visualized and understood. We are aiming for transparency and relying on the intelligence of contracting officers to deal with the fact that the data returned is likely highly imperfect.
We need a data set which can be made publicly available in order to stand up a demo instance of this site. If you have real, actual transactional data where some organization actual bought specific quantities of specific things at a specific date in time that you don't mind the whole world seeing, then please contact me: [email protected] and I will owe you deep gratitude.
Secondarily, we would like coding and design volunteers, not so much because we need the labor, but because I want this to become a living project outside of the Federal government that the government can draw upon for code, inspiration, and know-how.
Here is a preliminary TODO list:
- Factor out the PriceHistory specific GUI elements and fields in order to make the project both
cleaner and potentially applicable to other forms of data. - Factor out the GSA branding so that a different entity could comfortably use this project for its own purposes.
- Clean up my not very good javascript, primarily by factoring it into separate files that can be loaded.
- Help make the GUI much better than it is today. We are activley working on this, so there is of course the danger of some duplication of effort---but so be it.
- We need an export mechanism of portfolios (or the page) into Word, Excel, and PDF. Today there is an export tool that products very simple HTML suitable to recording a portforlio for research purposes. However, our uses have also asked for direct Excel and Word export.
Our goal is to save the government a lot of money in 2014. Although ambitious, if we need only make government buyers a few perecentage points more efficient to accomplish this. Your help may make all the difference in the world---and may easily exceed our own efforts.
Please direct questions to Robert L. Read at [email protected].
Thanks to John Hardy for attempting to work through some of this.
These instructions are probably insufficient for a clean install right now. I will improve them as soon as I can.
The PriceHistoryGUI project is separated from its sister, PriceHistoryAPI, to keep a nice clean distinction between the GUI and Middle (Business Logic) Layer. In practice, a user probably wants to install the PriceHistoryAPI first. Additionally, it is dependent on a third project of mine, the MorrisDataDecorator.
I've been told I should be able to get all of this working with PyPi, but until I do, here is an attempt at installation instructions.
You will need to have installed:
- Python
- easy_install
- bottle
- requests
- SOLR
- solrpy-0.9.5 [Note: this does not insall with PIP. I found "sudo python setup.py install" worked.] [Additional Note: Recently (in December of 2013) a new version of solrpy (0.9.6) was released. I am in the process of testing this module. It DOES install easily from PyPI.]
I use Apache, mod_wsgi, and mod_ssl, but mod_ssl is optional.
You will also need some data. I have provided an example data file in the "cookedData" directory in PriceHistoryAPI, but it only contains a few transactions and is not interesting. I am working with the government to get a proper file released, but have so far not been able to get around security and privacy concerns. However, there is a documented process for building an adapter to your own Prices Paid data file which you can follow.
Additionally, this project takes advantage of the excellent javascript projects gifted to the world:
- jquery
- jquery-ui
- jqplot
- SlickGrid (Huge thanks to Michael Leibman!)
- feedback_me (Thanks, Daniel!)
- jqPagination (Thanks, Ben Everard!)
Although in general configuration files allow flexibility in how to organize the system, I'm currently using a single directory for the install of the MorrisDataDecorator, the PriceHistoryAPI, and the PriceHistoryGUI.
Here are the recommended steps:
- Install the MorrisDataDecorator following the instructions found there. This project has automated unit tests. It implements a website, which you might wish to briefly bring up in bottle to be assured that it works, but really we are using it of the API definition and some back-end implementation.
- Create a "logs" directory parallel to MorrisDataDecorator, PriceHistoryAPI, and PriceHistoryGUI.
- Create an empty __init__.py file parallel to MorrisDataDecorator, PriceHistoryAPI, and PriceHistoryGUI.
- Install the PriceHistoryAPI following the instructions found there.
- Install the PriceHistoryGUI following the instructions found
- Install P3Auth.
- Create a "configuration" directory parallel to P3Auth.
- In P3Auth, edit "AuthBuilder" to set your own salt in P3APISALT. This will be a secret that you keep private and will also set in your apache environment variables. Then edit this line : " username = "changeme"+str(i) " in order to set your own username (possibly you will want to complete change this mechanism.) Then execute authbuilder to generate randome passwords and the hashes for them in the configuration directory. You will need to use at least one of these for accessing the API.
- Install the MorridDataDecorator in a directory parallel to PriceHistoryGUI and PriceHistoryAPI.
- Running the tests in the MorrisDataDecorator might be enlightening.
- Create a directory called "logs" parallel to all of those.
- In the logs directory, execute the "bash ../P3Auth/CreateLogFiles.bash"
- Create a directory called "js" which will be parallel to all of these previous directories. Configure Apache to allow files to be read from this directory freely---nothing secure will go here.
- In js, install jquery, jquery-ui, feedback-me, excanvas.js, jqplot, and jqPaginate. These are all third-party projects that I have been unwilling to duplicate into these repositories. I know this makes it very difficult to install---I am seeking a solution to this problem.
- Now, parallel to "js" install the project "SlickGrid-master" from SlickGrid. Also make this freely readable by anone hitting your site. The SlickGrid project is truly excellent.
- Copy the the file in PriceHistoryGUI/docs/Example.ppGuiConfig.py into PriceHistoryGUI/ppGuiConfig.py and edit it appropriately.
- Copy the file morris_config.py into
- Create your own "cookedData" directory with my example or your own with your own adapter. Before you create your own data, you may wish to use the file FY14TX-pppifver-USASpending-5-0-0-0-1.csv. This file is simply a renamed (and unchanged) export from the site USASpending.gov, in this case for fiscal year 2014 and the state of Texas. You will find it in the PriceHistoryAPI project in the example cookedData directory there.
- Carefully adjust ppGuiConfig.py to match your installation instructions.
- Install SOLR.
- In the PriceHistoryAPI directory, execute "python SolrLodr.py" to load the data in your cookedData directory into SOLR. Use the Solr administrative interface to make sure the documents are correctly inject. SolrLodr produces a log file of errors.
- Start up Apache and try to get it working. ** Start by using the exampe_apache.config in PriceHistoryAPI/docs. This is an example httpd.conf that should be really useful because it is based on real example. ** You will of course have to change many things in this file. In particular note that the the python-path variable sent to WSGI is very important.
- Set Environment variables similar to the following, possible in the profile for the user that runs apache or in the apache httpd.conf: ** SetEnv P3APISALT CHANGEME (use the salt you used above!) ** SetEnv PriceHistoryAPIUsername CHANGEME (use a username/password generated with authbuilder.) ** SetEnv PriceHistoryAPIPassword CHANGEME (use a username/password generated with authbuilder.) ** SetEnv PYCAS_SECRET PutANiceLongSecretHere
- In the MorrisDataDecorator, execute "run_all.bash" to start up Botlle processes listening on certain ports,
- Restart Apache and browse to the website.
- Use one of the username/password pairs generated by Authbuilder.py
At this point, probably after some configuration troubleshooting described below, you will have an empty site. To load data into the site, you need to load the SOLR index. This requires three steps:
- Copying the exampe scehma.xml into the correct position, and
- Creating a directory, by default called "cookedData" that is parallel to your other directoris, which contain comma-separated value files obeying the documented file naming convention, and
- Execute in PriceHistoryAPI "python SolrLodr.py", which inserts the transactions in the "cookedData" directory into the SOLR index. It runs in chunks of a 1000 at a rate of about one million transactions per hour.
In PriceHistoryAPI you will find a directory called "cookedData" which contains a file containing exactly 2 fake records. We hope to get some better data soon, but that is enough for you to exercise the system. This entire directory needs to be copied up to be parallel to PriceHistoryGUI.
We use a fresh, out-of-the-box instal of SOLR. In recent releases, the location of the schema.xml file is:
solr-4.4.0/solr/example/solr/collection1/conf/schema.xml
However, SOLR is a fast evolving project and there are different distributions that install in defferent ways. The most important thing is to add the definitions of fields in the file "schema.SECTION.xml" must exist in the schema.xml file. I like to put these after the definition of the "id" field, but I am not an SOLR expert.
Copy the file "schema.xml.example" in PriceHistoryAPI into this location and restart SOLR.
When you have done that and set up your cookedData directory, you are ready to execute "python SolrLodr.py".
A lot can go wrong in the installation of P3. The best thing to do is to email me ([email protected]) immediately. I will guide you through using the log files to figure out what is wrong. There are three basic log files that will be invaluable to you:
- The apache error and access log files,
- The SOLR output logs, which you will be responsible for, depending on how you start solr (I just pipe the out put of "java -jar start.jar" into a known location and tail that), and
- The P3 application log file which is by default in the logs directory and named Activity.log.
Generally you will use those files in that order.
In practice to do some debugging I have had to:
- Replace app.wsgi with a "hello world" version,
- Print out enviornment variables in app.wsgi,
- Generally do all the things you have to do to get Apache to correctly render a wsgi app,
- Look for coding mistakes that show up in the Apache error log,
- Look for misconfiguration that show up in the Apache error log,
- Look at the SOLR log to tryin to understand what is going on,
- Use the Activity.log file, sometimes with additional debugging statements, to understand what is going wrong there, and
- Use the "*.out" files in the MorrisDataDecorator directory on occation.