[HUD] Free, open source poker tracking software: FPDB - (Version 0.40.5) - Nov 14, 2013
10-12-2008
, 05:33 PM
A few more questions 
Anyone know of any nice IDE type editors, that will help me cross reference what code goes to where, and which does what when?
Is there a document describing the PS hand history format, and/or the eventual "fpdb" format?
Since I currently do all my play on Everleaf and Cake, once I understand what all is going on in here, I guess I might be the one leading the charge to get alternative site support
Anyone know of any nice IDE type editors, that will help me cross reference what code goes to where, and which does what when?
Is there a document describing the PS hand history format, and/or the eventual "fpdb" format?
Since I currently do all my play on Everleaf and Cake, once I understand what all is going on in here, I guess I might be the one leading the charge to get alternative site support
10-13-2008
, 09:30 PM
My two cents on possible places for performance improvement:
- As noted, many places of the python code aren't exactly well written or performant. However, the VAST majority of CPU load is in MySQL - when bulk importing I have around 85-90% MySQL load (on small DBs, 2-5k hands). So my best guesses for places for performance improvements in import are:
- If possible, commit multiple lines of e.g. HandsActions in one go, rather than one SQL access per line.
- Use a unix socket (or the Windows equivalent, assuming there is one) for accessing the DB rather than TCP/IP.
- Using PostgreSQL or SQLite may or may not produce significantly better or worse performance
- Remove useless SQL accesses (e.g. siteID could be hardcoded easily as it is filled by table setup anyways) and change crappy ones (e.g. I'm not buying that it is necessary to do a SELECT after an INSERT just to get the id of the new field...).
- using
might help a bit.
However, quite honestly, I think the biggest performance gains could be made by optimising the database itself, e.g. ensuring it's on a fast drive (or at least in the fastest region of whatever drive(s) you have), using RAID0, 5 or 6 if you can, fiddling with configuration options of the DB such as caching, ...
Now onto some other things
Actually the HUD can display much more, you just need to edit the rows and columns fields then add lines - e.g. I have 20 fields for holdem, and default for everything else.
The dependencies (listed on the webpage) and a text editor. Though ideally you should use git to always grab the latest version of my tree, but if you really hate git just use the latest alpha release.
To send patches:
- Ideal would be for you to have a public git tree
- Alternatively I'll take diffs
- As a last resort I'd take whole files.
On python exception handling: I honestly don't know if it's faster than "if divider!=0" or not (I'd be inclined to agree with you, a simple numerical check is pretty much by definition faster than exceptions. fpdb_import now has some timing, you could use that as an example. However, as noted above, I don't think that too much can be gained there, tho every tick counts and assuming there is no reduction in reliability or cleanness I'd take any performance improving patch.
Editors: My best suggestion is Exclipse+pydev, from your posts I'm guessing you have a sufficiently powerful machine to use those. There are others but I can't remember the names (one might've been called SPE), I tried one or two and was rather disappointed tbh. But I'm not a big IDE fan so don't take my word for it.
Format: No formal spec. However, there's examples in the form of real PS hands in the regression-test folder.
Sites: Looks that way
If their format is similiar to PS or FTP then I think the easiest&best way to do this would be to write a converter that turns your cake/everleaf hands into PS or FTP hands and have that called if such a history file is detected.
Is there no file fpdb-error.log or fpdb-error-log.txt?
That shouldn't work anymore, significant errors should be stored in a log file. (this will at some point be fixed so that the errors are stored in the log as well as printed to stdout) Note that the file may not be filled until fpdb exits, though that's obviously not the issue for you since it exits immediately.
Brilliant, keep us posted 
Instructions would be very useful, afaik there is no tracker that runs in mac.
- As noted, many places of the python code aren't exactly well written or performant. However, the VAST majority of CPU load is in MySQL - when bulk importing I have around 85-90% MySQL load (on small DBs, 2-5k hands). So my best guesses for places for performance improvements in import are:
- If possible, commit multiple lines of e.g. HandsActions in one go, rather than one SQL access per line.
- Use a unix socket (or the Windows equivalent, assuming there is one) for accessing the DB rather than TCP/IP.
- Using PostgreSQL or SQLite may or may not produce significantly better or worse performance
- Remove useless SQL accesses (e.g. siteID could be hardcoded easily as it is filled by table setup anyways) and change crappy ones (e.g. I'm not buying that it is necessary to do a SELECT after an INSERT just to get the id of the new field...).
- using
Code:
UPDATE HudCache SET vpip = vpip+1, prf = pfr+0, ... WHERE ...;
However, quite honestly, I think the biggest performance gains could be made by optimising the database itself, e.g. ensuring it's on a fast drive (or at least in the fastest region of whatever drive(s) you have), using RAID0, 5 or 6 if you can, fiddling with configuration options of the DB such as caching, ...
Now onto some other things
To send patches:
- Ideal would be for you to have a public git tree
- Alternatively I'll take diffs
- As a last resort I'd take whole files.
On python exception handling: I honestly don't know if it's faster than "if divider!=0" or not (I'd be inclined to agree with you, a simple numerical check is pretty much by definition faster than exceptions. fpdb_import now has some timing, you could use that as an example. However, as noted above, I don't think that too much can be gained there, tho every tick counts and assuming there is no reduction in reliability or cleanness I'd take any performance improving patch.
Quote:
Anyone know of any nice IDE type editors, that will help me cross reference what code goes to where, and which does what when?
Is there a document describing the PS hand history format, and/or the eventual "fpdb" format?
Since I currently do all my play on Everleaf and Cake, once I understand what all is going on in here, I guess I might be the one leading the charge to get alternative site support
Is there a document describing the PS hand history format, and/or the eventual "fpdb" format?
Since I currently do all my play on Everleaf and Cake, once I understand what all is going on in here, I guess I might be the one leading the charge to get alternative site support
Format: No formal spec. However, there's examples in the form of real PS hands in the regression-test folder.
Sites: Looks that way
If their format is similiar to PS or FTP then I think the easiest&best way to do this would be to write a converter that turns your cake/everleaf hands into PS or FTP hands and have that called if such a history file is detected.
Quote:
What should I -expect- to happen when running fpdb.py ?
Here's what I actually get:
no error logs.
Here's what I actually get:
Code:
E:\fpdb\pyfpdb>fpdb.py Failed to load libs for graphing, graphing will not function. Please install num py and matplotlib. todo: implement obtain_global_lock Opened and read profile file C:\Documents and Settings\Eric.JUBY\Application Dat a\fpdb\default.conf E:\fpdb\pyfpdb>
Instructions would be very useful, afaik there is no tracker that runs in mac.
10-13-2008
, 10:40 PM
I've managed to get fpdb running without a lot of excessive hacking today. Despite getting the error that numpy and matplotlib couldn't be loaded, I can get a simple graph up. Loading my HH is the extent of my current tests. Testing the table viewer and HUD will have to wait a couple of days.
I've attached some notes below, feel free to reproduce them in any way that you'd like. These steps actually use the X server, but I think some additional configuration will remove the need for it.
Quote:
0. Need XCode 2.5 or better and some version of X windows. I'm using macports for as much as possible. Head to
http://www.macports.org/
and pick up a dmg for your OS X version.
NOTE: MacPorts will download source for all packages and compile binaries. Depending on your computer, connection etc, the steps below will take several hours to complete. Your complete attention will not be needed for most of it though.
1. Macports for python, pygtk, etc
Get python 2.5:
sudo port install python25
sudo port install python_select
sudo python_select python25
Now we want pygtk.
sudo port install py25-gtk
will install glib2 and the other dependencies. We also want numpy and matplotlib
sudo port install py25-matplotlib
Now mysql-python
sudo port install py25-mysql
This installs mysql5 as a dependency.
2. Configure mysql
Some info on configuring mysql can be found at
http://2tbsp.com/content/install_and...sql_5_macports
sudo /opt/local/lib/mysql5/bin/mysql_install_db --user=mysql
sudo mkdir /opt/local/var/run
sudo /opt/local/bin/mysqld_safe5 &
sudo /opt/local/lib/mysql5/bin/mysql_secure_installation
Now run
mysql5 -u root -p
CREATE DATABASE fpdb;
GRANT ALL PRIVILEGES ON fpdb.* TO 'fpdb'@'localhost' IDENTIFIED BY 'newPassword' WITH GRANT OPTION;
Replace newPassword with a suitable password.
3. Configure fpdb
Now create ~/.fpdb and copy docs/default.conf to it. Edit the file and enter the password you set in mysql just above.
Now start your X windows client, open a shell, find the pyfpdb/fpdb.py file and run
pythonw path_to/fpdb.py &
http://www.macports.org/
and pick up a dmg for your OS X version.
NOTE: MacPorts will download source for all packages and compile binaries. Depending on your computer, connection etc, the steps below will take several hours to complete. Your complete attention will not be needed for most of it though.
1. Macports for python, pygtk, etc
Get python 2.5:
sudo port install python25
sudo port install python_select
sudo python_select python25
Now we want pygtk.
sudo port install py25-gtk
will install glib2 and the other dependencies. We also want numpy and matplotlib
sudo port install py25-matplotlib
Now mysql-python
sudo port install py25-mysql
This installs mysql5 as a dependency.
2. Configure mysql
Some info on configuring mysql can be found at
http://2tbsp.com/content/install_and...sql_5_macports
sudo /opt/local/lib/mysql5/bin/mysql_install_db --user=mysql
sudo mkdir /opt/local/var/run
sudo /opt/local/bin/mysqld_safe5 &
sudo /opt/local/lib/mysql5/bin/mysql_secure_installation
Now run
mysql5 -u root -p
CREATE DATABASE fpdb;
GRANT ALL PRIVILEGES ON fpdb.* TO 'fpdb'@'localhost' IDENTIFIED BY 'newPassword' WITH GRANT OPTION;
Replace newPassword with a suitable password.
3. Configure fpdb
Now create ~/.fpdb and copy docs/default.conf to it. Edit the file and enter the password you set in mysql just above.
Now start your X windows client, open a shell, find the pyfpdb/fpdb.py file and run
pythonw path_to/fpdb.py &