Kurt's Weblog

Val mentioned a program called elog, which is a very interesting little web application.

We started putting together a fink package, but it is quite a ways away from working. http://schwehr.org/software/fink/elog/

Just tried my first server size move of a file through subversion:

$ svn mv https://cowfish.unh.edu/projects/seismic-py/trunk/segy.py \
       https://cowfish.unh.edu/projects/seismic-py/trunk/seismic/segy.py



Committed revision 337.



$ cd && svn update
A    seismic/segy.py
D    segy.py
Updated to revision 337.

I would like to do something like this for ccom a soon as I get a server for it.

http://sidbox.homelinux.org/

This is where I put my compiled debs for future use when I tinker with fink.
If you are using fink/10.4-transitional you could save some times using my repositories, just add:



deb http://sidbox.homelinux.org/fink stable main crypto
deb http://sidbox.homelinux.org/fink unstable main crypto



at the end of your /sw/etc/apt/sources.list and:



UseBinaryDist: true



into your /sw/etc/fink.conf

There is also http://www.gtlib.gatech.edu/pub/fink/fink/

Here is a quick example to go with the previous on. I am sure this is not the best way to do it, but it works. For each row in the database, I want to print out all the fields with their values.

results = usermapper.select_text('select * from msg1 where mmsi=353556000 order by messageId')



for r in results:
    for attribute in r._managed_attributes.keys():
	print attribute,':     ',r.__getattribute__(attribute)

Question: What exactly is the difference between the __repr__ and __str__ class methods?

I figured out how to do the object relational mapping so that database comes back as a list of objects. A couple points about this code. First, I had to change messageId to be the primary_key. Second. The online docs have a mistake. It should not be "results=mapper.select_text" What I have below works. Third, printing just r gives things like '<__main__.Msg1 object at 0x1009970>'. I probably need to add a __repr__ to make print nice.

#!/usr/bin/env python
import os,sys
from sqlalchemy import *
db = create_engine('sqlite://filename=ais.db' )
msg1table = Table('msg1', db,
		  #  unique index
		  Column('messageId', Integer,primary_key = True)
		  ,Column('mmsi',Integer, index=True)
		  ,Column('lon',Float)
		  ,Column('lat',Float)
		  ,Column('SOG',Float)
		  ,Column('ROT',Float)
		  ,Column('time',TIMESTAMP)
		  )
class Msg1(object):
    pass # Yikes
msg1mapper = mapper(Msg1,msg1table)
results = msg1mapper.select_text('select * from msg1 where mmsi=353556000 order by messageId')
for r in results:
    print r.lon, r.lat

Running this gives:

PYTHONPATH=.. ./ais-sql2geozui 
-70.7583333333 43.08
-70.7583166667 43.0799833333
-70.7583166667 43.08
-70.7583 43.08
-70.7583 43.0799833333
-70.7583 43.0799833333

I've hit a roadblock at the moment trying to build openEV in fink. I am using a cvs checkout from today of OpenEV for the build. /sw/include/glib-2.0/gobject/gtype.h is giving me trouble here.

#if	GLIB_SIZEOF_LONG == GLIB_SIZEOF_SIZE_T
typedef gulong                          GType;
#else	/* hm, shouldn't happen? */
typedef gsize                           GType;
#endif

Here is my current configure scripts:

CFLAGS="-I/sw/include/gtk-1.2 -I/sw/include/glib-1.2 -I/sw/lib/glib/include \
-I/usr/X11R6/include -I/sw/include/pygtk-2.0 -I/sw/include/gdal1 \
-I/sw/include" LDFLAGS="-L/sw/lib" \
./configure --prefix=/Users/schwehr/Desktop/openev 

Here is a little sioseis script from Becca. This does not work for all datasets. For example, my Humboldt Slide dataset was not correctable in this way.

First her intro:

I have got the heave correction to work for wire in-out issues where
you can't see the multiple! I took the depth profile from fledermaus
and converted it to twtt and used it as a picked water bottom. To do
this I imported the xy data points that I used to make the vertical
curtain and draped them on the seafloor, made a note of the depth of
each one and then converted that depth into twt (using 1500 m/s), Then
I made a file with the shot number and the fleder seafloor depth and
used that in the script below. It only works if there is not
significent layback.



-Becca

And the script:

pre>
#!/usr/bin/env bash



# for shifting when you've picked the multiple/know the seafloor depth or time

declare -r FNO="94000"
declare -r LNO="111047"
declare -r BASENAME=ice2003_B7



sioseis << eof
procs diskin header mix wbt wbt2 header2 prout shift diskoa end



diskin
   set 0 .5
   ipath $BASENAME.sgy end
end



header
    fno $FNO lnO $LNO ftr 1 ltr 99 l6 = l3 end
end



mix
    type 4  lhdr 59 weight
    1 1 1 1 1 1 1 1 1 1
    1 1 1 1 1 1 1 1 1 1
    1 1 1 1 1 1 1 1 1 1 end
end



wbt
! shot-time pairs for fleder picked seafloor gets asigned to r50
`cat B7.sf`
end



wbt2
! pick the water bottom from the original data, using the method that works best beneath
   index 60   ! put this pick in real word 60



!   peak pos sepp .07 .09 track .01 lprint 2 end
!   thres .1e-08 track .005 end
!   peak pos sepp 0.01 0.25 track 0.005 lprint 2 end
    solrat 6 ses 0.0010 0.0020 sel 0.1 0.25 lprint 2 end
!   solrat 4 ses .001 .005 sel .2 .3 lprint 2 end
end



header2
   fno $FNO lno $LNO ftr 1 ltr 99
      ! substitute the picked water column depth for the theoretical
   r59 = r50 - r60 end
end
prout
   indices r50 r59 r60
   fno $FNO lno $LNO noinc 100 end
end
shift
  indices r59 end
end
diskoa
   ofmt 1 opath $BASENAME-flt.sgy end
end
end
eof

So now on to inserting some data into the database with sqlalchemy. It does not look like a list of dictionaries is possible, but this is not so bad. This example is a little complicated because I am translating between the ais VDM module names and what I might want to call things in the database. Who wants to write out Longitude all the time when 'lon' is pretty well known?

aiskeys = [
    ['UserID','mmsi']
    ,['Longitude','lon']
    ,['Latitude','lat']
    ,['SOG','SOG']
    ,['ROT','ROT']
]
for key in aiskeys:
    insdict[key[1]] = fields[key[0]]
insdict['time'] = ais.nmeamessages.zdaDict2TIMESTAMP(timeDict)
msg1table.insert().execute(insdict)

The insertion dictionary looks like this:

insdict = 
  {
  'mmsi': 353556000, 
  'lon': -70.75833333333334, 
  'lat': 43.079999999999998, 
  'SOG': 0, 
  'ROT': 0,
  'time': '2006-03-26 13:39:21'
  }

Then I can pull data on the command line:

sqlite ais.db 'select * from msg1' | tr '|' '\t'
        353556000       -70.7583333333  43.08           0       0       2006-03-26 13:39:21
        353556000       -70.7583166667  43.0799833333   0       0       2006-03-26 13:39:31
        224632000       -70.7909966667  43.103435       1       128     2006-03-26 13:39:32
        353556000       -70.7583166667  43.08           0       0       2006-03-26 13:39:42
        353556000       -70.7583        43.08           0       0       2006-03-26 13:39:51
        224632000       -70.7909933333  43.10343        1       128     2006-03-26 13:40:03
        224632000       -70.7909933333  43.10343        1       128     2006-03-26 13:40:03
        224632000       -70.7909933333  43.10343        1       128     2006-03-26 13:40:03
        353556000       -70.7583        43.0799833333   0       0       2006-03-26 13:40:11
...

Wow, that was a little complicated for a first example, but it was so easy to write.

A quick look in the sqlite driver within sqlalchemy show that it is using the pysqlite2 driver (for sqlite3) and if not, it falls back on pysqlite which is will only generate sqlite2 databases. sqlite2 is pretty clunky, so I need to get the pysqlite2 drivers into fink soon! Here is how they do conditional driver loading. An interesting read:

try:
    from pysqlite2 import dbapi2 as sqlite
except:
    try:
        sqlite = __import__('sqlite') # skip ourselves
    except:
        sqlite = None

Now here is my first attempt at creating a database via sqlalchemy:

#!/usr/bin/env python
'''Use SQLAlchemy to create/add data to a database'''
from sqlalchemy import *
import os
if os.access('ais.db',os.W_OK):
    os.remove('ais.db')
#
db = create_engine('sqlite://filename=ais.db',echo=True)
#
msg1table = Table('msg1', db,
		  #  unique index
		  Column('messageId', Integer,unique=True)
		  ,Column('mmsi',Integer, index=True)
		  ,Column('lon',Float)
		  ,Column('lat',Float)
		  ,Column('SOG',Float)
		  ,Column('ROT',Float)
		  ,Column('time',TIMESTAMP)
		  ).create()

Running it looks like this. The key is the "echo=True".

./ais-sqlalchemy
[2006-03-30 11:18:44,781] [engine]: 
CREATE TABLE msg1(
	messageId INTEGER, 
	mmsi INTEGER, 
	lon NUMERIC(10, 2), 
	lat NUMERIC(10, 2), 
	SOG NUMERIC(10, 2), 
	ROT NUMERIC(10, 2), 
	time TIMESTAMP
)






[2006-03-30 11:18:44,803] [engine]: None
[2006-03-30 11:18:44,838] [engine]: CREATE UNIQUE INDEX ux_messageId ON msg1 (messageId)
[2006-03-30 11:18:44,838] [engine]: None
[2006-03-30 11:18:44,842] [engine]: CREATE INDEX ix_mmsi ON msg1 (mmsi)
[2006-03-30 11:18:44,842] [engine]: None

Since I don't have the newer driver, the above code generates and sqlite2 database.

less ais.db



** This file contains an SQLite 2.1 database **^@
...



sqlite ais.db '.tables'
msg1



sqlite ais.db '.schema msg1'
CREATE TABLE msg1(
        messageId INTEGER, 
        mmsi INTEGER, 
        lon NUMERIC(10, 2), 
        lat NUMERIC(10, 2), 
        SOG NUMERIC(10, 2), 
        ROT NUMERIC(10, 2), 
        time TIMESTAMP
);
CREATE INDEX ix_mmsi ON msg1 (mmsi);
CREATE UNIQUE INDEX ux_messageId ON msg1 (messageId);

I have just started to play with SQLAlchemy now that I have working fink packages for setuptools and SQLAlchemy. I promise will check them in when the sf.net CVS server comes back online. In the process, I am working on creating a db of AIS data. In ipython I wanted some more info on various time functionality and I got quite a surpise.

In [23]: sqlalchemy.Time?
Type:           type
Base Class:     <type 'type'>
String Form:    <class 'sqlalchemy.types.Time'>
Namespace:      Interactive
File:           /sw/lib/python2.4/site-packages/sqlalchemy/types.py
Docstring:
    <no docstring>



In [24]: help(sqlalchemy.Time)



Help on class Time in module sqlalchemy.types:



class Time(TypeEngine)
 |  Method resolution order:
 |      Time
 |      TypeEngine
 |      __builtin__.object
 |  
 |  Methods inherited from TypeEngine:
 |  
 |  __init__(self, *args, **kwargs)
 |  
 |  adapt_args(self)
...

No doc string? I figured I could go look at the source and see what is goind on. Maybe the doc string was getting nuked somehow? Nope! I found this in /sw/lib/python2.4/site-packages/sqlalchemy/types.py:

...
class DateTime(TypeEngine):
    pass



class Date(TypeEngine):
    pass



class Time(TypeEngine):
    pass



class Binary(TypeEngine):
    def __init__(self, length=None):
        self.length = length
    def convert_bind_param(self, value, engine):
        return engine.dbapi().Binary(value)
    def convert_result_value(self, value, engine):
        return value
    def get_constructor_args(self):
        return {'length':self.length}
...

Say what? First off, there is not inline documentation or comments. Secondly, a class with just pass? So this will inherit the parents functionality, but appear as a different object. Is that the point?

Randomness of the web. I was looking for that sf.net status page since they appear to be having a cvs outage when I ran into OpenSounding:

  A tool for decoding and presenting Geotechnical Sounding data (such
  as CPTu). Formats supported are the SGF standard as defined by the
  Swedish Geotechnical Society (www.sgf.net), and used by a number of
  commercial products, aswell as Generic text

I think I found the trouble for setup tools install from fink.... /sw/lib/python2.4/site-packages/easy-install.pth was:

/sw/src/fink.build/root-setuptools-py24-0.6a10-1/sw/lib/python2.4/site-packages\
/setuptools-0.6a10-py2.4.egg

Which should not have all that stuff in the beginning:

/sw/lib/python2.4/site-packages/setuptools-0.6a10-py2.4.egg

Brian L mentioned Cruise Control which is a continuous build process system. Sounds interesting. I hadn't gotten into BuildBot yet. I need to evaluate the two.

WorldWind2D (WW2D). Check out the screen shot.

I am starting to have strange problems with X11. Often xterms popup and disappear. Then when they have been iconofied, the come back looking like this. It would be a lot easier to track down if these were consistant problems, but they only happen occasionally.

From http://www.openscience.org/blog/. How to make money from Open Source scientific software

...
  However, their code is not open. And this, I think, is a real
  problem. Imagine a skeptical researcher who is sent a paper to
  review. Further imagine that this paper uses this company's
  software, and the skeptical researcher doesn't have the money for a
  license. He or she therefore can't "look under the hood" to verify
  what's going on if they have some questions about how the code is
  calculating something relevant to the paper. There are good reasons
  that this company doesn;t give away their code; they like to put
  food on the table, and they don't trust the rest of the community to
  shell out the money for their programs if the code were available
  for free.



  So, I'm left with a dilemma. I want this company to do well, to hire
  more of my students in the future, and to continue to produce high
  quality code. I also want the codes that we use in my field to be
  available for skeptical review. So today, I'm starting a set of
  posts in which I'll try to hash out the following question: How can
  people make money from open source scientific software?
...

http://www.perrygeo.net/wordpress/?p=35

  Web Mapping Services (WMS) are not always my prefered option for
  accessing data; relying on a remote server to generate a pretty
  picture of the data is hardly a substitute for having the raw data
  in hand. But for many cases, I just need a decent looking basemap
  image and don't want to download gigabytes of data, especially if
  that data is updated frequently.
...

To play with setuptools, I has to do a little extra work. I downloaded the zip from the cheeseshop and did this.

cd ~/Desktop
mkdir setuptools && cd setuptools
wget http://cheeseshop.python.org/packages/source/s/setuptools/setuptools-0.6a10.zip



mkdir -p /Users/schwehr/Desktop/setuptools/sw/lib/python2.4/site-packages



PYTHONPATH=/Users/schwehr/Desktop/setuptools/sw/lib/python2.4/site-packages/ \
 python setup.py install --root=/Users/schwehr/Desktop/setuptools/

If I did not put the target site-packages location in PYTHONPATH it failed to install.

Here is the port file for py-setyptools: cat /opt/local/var/db/dports/sources/rsync.rsync.darwinports.org_dpupdate_dports/python/py-setuptools/Portfile

# $Id: Portfile,v 1.1 2005/12/15 00:55:31 blb Exp $



PortSystem			1.0
PortGroup			python24 1.0
name				py-setuptools
version				0.6a8
categories-append	devel
maintainers			blb@opendarwin.org
description			distutils enhancement for build and distribution
long_description \
	setuptools is a collection of enhancements to the Python distutils that \
	allow you to more easily build and distribute Python packages, \
	especially ones that have dependencies on other packages.



platforms			darwin



homepage			http://peak.telecommunity.com/DevCenter/setuptools
master_sites		http://cheeseshop.python.org/packages/source/s/setuptools/
distname			setuptools-${version}
use_zip				yes



checksums			md5 3eecdf66c1a2cf8a6556bc00b69d572a



post-destroot {
	xinstall -m 755 -d ${destroot}/${prefix}/share/doc/${name}
	xinstall -m 644 -W ${worksrcpath} EasyInstall.txt api_tests.txt \
		pkg_resources.txt setuptools.txt \
		${destroot}/${prefix}/share/doc/${name}
}

There are two very interesting looking database abstraction layers for python that both require setuptools. setuptools is not in fink, but it is in darwinports, so I am installing it that way to see how darwinports handles it. Hopefully after setuptools, I can go back and add wxoptparse to fink too! That would be great! I still haven't managed to get wxglade working yet. http://py-setuptools.darwinports.com/

  sudo port install py-setuptools

I should have added "-v -d" to the above.

A while back (Fall 2004), I made a presentation entitled "How to stop worrying and love the blob." Needless to say I missed the target with my audience completely, but now OpenBSD has a very similiar theme (but no Dr. Strangelove references).

I bought an OpenBSD CD. What are you doing to support the puffer fish?

Linux supporters fiddle while OpenSSH burns

http://openbsd.org/orders.html

Yes, I am having fun. I am only using sqlite and gnuplot so far, but I will keep stepping it up. The numbers in the key are the ship UserID/MMSI.

Thanks to Roland for this link.

"In 5-10 years, most of the software you use will be free"

  Greg Stein, chairman of the Apache Software Foundation, currently with
  Google, delivered this keynote on Wednesday morning at EclipseCon
  2006. It was titled "Comparing Apache and Eclipse".

A very interesting read! Could be very controversial to some. Quite the contrast to listening to the AVSP interviews with some folks at ESRI that I was listening to earlier today.

How do I correctly set the python interpreter using setup.py for scripts that end up in /sw/bin and what about name clashes? Guess I only allow one version of python? That just does not seem right. Plus, what I am doing now is not safe:

#!/usr/bin/env python

Then what happens when the user changes their path to pull some other python before the one that I am counting on? distutils must have thought about this, no?

This not the most clever implementation of twos complement signed integer decoding, but it seems to work. I am putting this here since Avi is holding off on additional functionality in BitVector until the end of his semester. He did fix the bitvec[-1]=0 reverse accessor bug today, so the code below could be even simpler.

#!/usr/bin/env python
from BitVector import BitVector



def setBitVectorSize(bv,size=8):
    """Pad a BitVector with 0's on the left until it is at least the size specified



    @param bv: BitVector that needs to meet a minimim size
    @type bv: BitVector
    @param size: Minimum number of bits to make the new BitVector
    @type size: int
    @return: BitVector that is size bits or larger
    @rtype: BitVector



    @todo: What to do if the vector is larger than size?
    """
    pad=BitVector(bitlist=[0])
    while len(bv)<size: bv = pad + bv
    return bv



def addone(bv):
    '''
    Add one bit to a bit vector.  Overflows are silently dropped.
    >>> print addone(BitVector(bitstring='1100'))
    1101
    >>> print addone(BitVector(bitstring='1111'))
    0000



    @param bv: Add one to these bits
    @type bv: BitVector
    @return: Bits with one added
    @rtype: BitVector
    '''
    new = bv
    r = range(1,len(bv)+1)
    for i in r:
	index = len(bv)-i
	if 0==bv[index]:
	    new[index]=1
	    break
	new[index]=0
    return new



def subone(bv):
    '''
    Subtract one bit from a bit vector



    >>> print subone(BitVector(bitstring='1111'))
    1110
    >>> print subone(BitVector(bitstring='0010'))
    0001
    >>> print subone(BitVector(bitstring='0000'))
    1111



    @param bv: Bits to add one bit to the right side
    @type bv: BitVector
    @rtype: BitVector
    '''
    new = bv
    r = range(1,len(bv)+1)
    for i in r:
	index = len(bv)-i
	if 1==bv[index]:
	    new[index]=0
	    break
	new[index]=1
    return new



def bvFromSignedInt(intVal,bitSize=None):
    '''
    Create a twos complement BitVector from a signed integer.



    Positives must have a '0' in the left hand position.



    >>> print bvFromSignedInt(0,bitSize=4)
    0000
    >>> print bvFromSignedInt(1,bitSize=4)
    0001
    >>> print bvFromSignedInt(7,bitSize=4)
    0111



    Negative numbers must have a '1' in the left hand position.



    >>> print bvFromSignedInt(-1,bitSize=4)
    1111
    >>> print bvFromSignedInt(-2,bitSize=4)
    1110
    >>> print bvFromSignedInt(-7,bitSize=4)
    1001



    @param bv: Bits to subtract one bit from the right side
    @type bv: BitVector
    @rtype: BitVector
    '''
    bv = setBitVectorSize(BitVector(intVal=abs(intVal)),bitSize-1)
    assert(bitSize-1==len(bv)) # FIX: what to do about overflow beyond bitSize?
    if intVal>=0:
	bv = BitVector(intVal=0) + bv
    else:
	bv = subone(bv)
	bv = ~bv
	bv = BitVector(intVal=1) + bv
    return bv
    
def signedIntFromBV(bv):
    '''
    Interpret a bit vector as an signed integer.  int(BitVector)
    defaults to treating the bits as an unsigned int.  Assumes twos
    complement representation.



    U{http://en.wikipedia.org/wiki/Twos_complement}



    Positive values decode like so:



    >>> signedIntFromBV(BitVector(bitstring='0000'))
    0
    >>> signedIntFromBV(BitVector(bitstring='0101'))
    5



    Here are some negative integer examples:



    >>> signedIntFromBV(BitVector(bitstring='1111'))
    -1
    >>> signedIntFromBV(BitVector(bitstring='1110'))
    -2
    >>> signedIntFromBV(BitVector(bitstring='1010'))
    -6
    >>> signedIntFromBV(BitVector(bitstring='1001'))
    -7



    @param bv: Bits to treat as an signed int
    @type bv: BitVector
    @return: Signed integer
    @rtype: int



    @note: Does not know the difference between byte orders.
    '''
    if 0==bv[0]: return int(bv)
    # Nope, so it is negative
    return -(int(addone(~(bv[1:]))))

This is not so exciting for those who have been using AIS for quite a while (e.g. Matt and Roland) but it is pretty exciting for me. First, this morning I created functions for the BitVector class that go to and from signed integers. That was pretty cool by itself, but that fix the last hurdle to fully decoding the longitude and latitudes from AIS messages 1-3. Those fields use signed integers of 10000 minute increments to store the location. Being in the western hemisphere, I was expecting negative longitudes (e.g. NOT 373 degrees).

To give ais-py a sanity test, I did a quick dump of the location and got these values for one message:

  MMSI/UserID Longitude Latitude SpeedOverGround
   353556000  -70.7583  43.07995        0

Google Maps/Local was kind enough to show me that this is indeed the salt hauler that was docked in Portsmouth over the weekend.

The CUAHSI () has a Hydrologic Information System that looks interesting. How do I access this data is a programmatic way (e.g. via python)?

Here is the USGS New Hampshire webpage. I am not too excited to do web scraping.

Of related interest: NASA/JPL's sensor webs project

Python 201 -- (Slightly) Advanced Python Topics by Dave Kuhlman looks pretty interesting. Some stuff is not how I do it in the testing section, but this is definitely worth a read.

How OS X Executes Applications by 0xFE. Looks like a good read and not too long.

pyserial is great! Here is how hard it was to create a very basic logging program.

#!/usr/bin/env python
import time
import serial



ser = serial.Serial('/dev/tty.KeySerial1',38400,timeout=2)
log=open('ais.log','a')
log.write('# START LOGGING AT GMTTIME: '+str(time.gmtime())+'\n')



i=0
while True:
    i += 1
    line = ser.readline()
    print i,len(line),':',line
    if len(line)>0:
	log.write('# GMTTIME: '+str(time.gmtime())+'\n')
	log.write(line)
	log.flush()

Making An Operating System Faster (10 Things Apple Did To Make Mac OS X Faster). An interesting read. Of note is how to make gcc precompiled headers for faster compiles.

% gcc -x c-header -c foo.h 
% ls foo.* 
foo.c foo.h foo.gch 

This is not bothering us, but it seems to be bothering some of the neighbors. The ships are not there that often. I would tend to side with the local shipping businesses. Yes, I can hear them unloading the salt right now.

Port Asked To Cut Overnight Work



PORTSMOUTH, N.H. -- Neighbors and a legislator are asking the Port of
New Hampshire in Portsmouth to stop overnight work so they can get
some sleep.
...

High-tech trackers: A local initiative to track vessels in and out of the Port of Hampton Roads is now being used in San Diego.

PORTSMOUTH -- A Coast Guard and Navy vessel tracking initiative that
began in Hampton Roads in 2001 is going national.



The Joint Harbor Operations Center, which monitors ships on local
waterways through a combination of radar systems, cameras and vessel
tracking devices, is now being implemented in San Diego.
...
"This center is about maintaining awareness," said Capt. Robert
O'Brien, the Coast Guard's captain of the port in Hampton Roads. "And
that awareness is key to the safety of our port."
...
The JHOC, as it's called, has evolved greatly from its early days
operating with only a handful of people out of the old watchtower on
the Elizabeth River at Norfolk Naval Station. Since the summer of
2004, the center has been headquartered at the Coast Guard base in
Portsmouth, with millions of dollars worth of computers, software and
equipment.
...
The idea for such a center is the brainchild of Capt. Joseph Bouchard,
the former commanding officer at the Norfolk Naval Station.
...

I just received my first AIS message. Not doing anything fancy. Just watching through minicom (installed via fink) using a KeySpan USA-19HS. It is set to 38.4k 8N1 looking at /dev/tty.KeySerial1. I just installed the latest KeySpan driver, so /dev/*USA* is no more.

        SR162 AIS Receiver                                                   
                                                                             
        RECE FRQ        R1F1619750                                           
        RECE FRQ        R2F1620250                                           
        INT RATE        I38400          
!AIVDM,1,1,,A,13F>Lh0P00JstMJHbOQskOvJ087h,0*5A
!AIVDM,2,1,1,A,v>W`lPt40030Ph0FhhQP1<v00C1W@8<4FS0?h000@P81O0P605P0PPm6,0*09
!AIVDM,2,2,1,A,P004P`4,2*75                                               
!AIVDM,1,1,,B,13F>Lh0P01JstLjHbOUrp?v80D00,0*46

I have to decode the messages, but this may be the Qing Ping Hai registered in Panama with an IMO number of 8308915. It took a few minutes to get all those messages.

I can't wait to get the RM808 back from service so I can see what it puts out.

Note: Do not unplug the keyspan while the port it is still open. That set off a kernal trap saying I had to use the power button to restart my laptop.

Mapping Heterogeneous Types with SQLAlchemy

Advanced Visual Effects with OpenGL [gamedev.net]

This is the suggestion for how to do line numbering that I got from Marine Geology.

Please find below a sample on how you will prepare your LaTeX document with line number:



\documentclass{elsart}
\usepackage{lineno}
\begin{document}
\linenumbers
\title{A Framework for Language Identification}
\author{John Smith, John Doe, John JohnJohn}
\address{Speech and Audio Research Laboratory,\\ Technology Way,\\ A City, AUSTRALIA, 4001.\\ \{jsmith, jdoe, johnjohnjohn\}@unversity.edu}
\begin{abstract}
Whilst several examples of segment based approaches to Language Identification (LID) have been published, they are not as good as this one.
\end{abstract}

Unfortunately that did not work for me. It mangled the title and abstract. I can not figure out how to no use \begin{frontmatter} which also stops the line numbering at the end of frontmatter. Also, the line number stops for the figure captions. Bummer. Here is the skeleton of what I did:

%\documentclass{elsart}
\documentclass[doublespacing]{elsart} % MARINE GEOLOGY DOUBLE SPACED
\usepackage{lineno}                   % LINE NUMBER PACKAGE
\usepackage{natbib}
\usepackage{amssymb}
\usepackage{graphicx}
\usepackage{array}
\usepackage{lscape} % landscape
\usepackage[usenames]{color} 



\begin{document}
\begin{frontmatter}
\title{Origin of continental margin morphology: submarine-slide or
  downslope current-controlled bedforms, a rock magnetic approach}
...
\end{frontmatter}
\linenumbers % Tell latex lineno package to start using line numbers
\section{Introduction}
Rest of the paper...

I hope that helps someone. I was not able to find this on the Elsevier web site. Their latex help page seems to have lost most of its content.

Nathan has setup 3 more public subversion projects for me.

• https://cowfish.unh.edu/projects/seismic-py/
• https://cowfish.unh.edu/projects/terrain/
• https://cowfish.unh.edu/projects/xsonar/

seismic-py represents a name change from segy-py to avoid a name clash (yes, I did have segy-py public first). No point in confusing people. I need to do freshmeat and cheeseshop (aka pypi) announcements this week.

xsonar is my protoized update to the USGS xsoanr sidescan. That means I converted K&R function prototypes to modern ANSI-C prototypes.

terrain is the same old terrain that tries to convert height field grids between a number of different formats (gmt grids, lightwave, Alias obj, OpenInventor, PDS, Vicar). Lightwave is a pretty crazy format, so my support for it has a lot of troubles.

http://www.worldwindcentral.com/wiki/Road_Map

This is good news. I don't mean that it is good that some things are not building, but it is good that there is a master list of things that are not building. There are a few of mine that are not building.

PPC: http://fink.opendarwin.org/build/2006-03-09/ppc/out/report.html

x86: http://fink.opendarwin.org/build/2006-03-09/i386/out/report.html

This sounds like a good thing to be able to do: Using Spotlight from the OS X Commandline

$ mdfind gmt.info



/Users/schwehr/Desktop/gmt.info.kds
/sw/fink/10.3/stable/main/finkinfo/sci/gmt.info
/sw/fink/10.4/stable/main/finkinfo/sci/gmt.info
/sw/fink/10.3/unstable/main/finkinfo/sci/gmt.info
/sw/fink/10.4/unstable/main/finkinfo/sci/gmt.info
/sw/fink/10.4-transitional/stable/main/finkinfo/sci/gmt.info
/sw/fink/10.4-transitional/unstable/main/finkinfo/sci/gmt.info
/Users/schwehr/Desktop/gmt.info
/Users/schwehr/projects/xcore/doc/backup.txt
...

Take a look at the screen shots on this web page of the Navionics Platinum here and here http://www.panbo.com/yae/archives/001119.html

http://www.navionics.com/PlatinumCharts.asp

http://www.seacoastonline.com/news/03222006/business/93865.htm

...
 Parker has trained as a lawyer and a ship captain. He's worked for
 more than three decades as a maritime legal expert and a yacht and
 ship broker. But Parker, a native of the fishing-famous town of
 Gloucester, Mass., said for years he's thought of creating a
 flexible, cost-effective and well-credentialed academy for both the
 recreational boating and commercial shipping markets. 
...
 Classes will begin April 17 at the AMA offices on Brewery Lane just
 off Islington Street. Parker said he plans to start small by
 offering basic U.S. Coast Guard certification classes.  
...

This is exciting. Nathan just migrated the first of my projects to subversion. This one is under the GPL, so I am allowing full subversion read access. Check it out at:

https://cowfish.unh.edu/projects/density

I passed Nathan a dump file that I generated from cvs2svn like this:

cvs2svn --dump-only --dumpfile ../../density.svndump density

Inside the density subdir from within my cvs repository were these files:

AMScrunch.f,v          VolHeader.C,v                 s_bootstrap.C,v
AUTHOR,v               VolHeader.H,v                 s_bootstrap.help2man,v
Attic                  as1-crypt.s,v                 s_bootstrap_cmd.ggo.in,v
Bootstrap.C,v          as2-slump.s,v                 sample.wpt,v
Bootstrap.H,v          as2xyz.bash,v                 simpleview.C,v
Cdf.C,v                as3-undef.s,v                 simpleview.help2man.in,v
Cdf.H,v                axes.iv,v                     simpleview.in,v
ChangeLog,v            bootvol-thesis.bash,v         simpleview_cmd.ggo.in,v
Density.C,v            debug.H,v                     spin_gnuplot.C,v
Density.H,v            debug.bash,v                  spin_gnuplot_cmd.ggo.in,v
DensityFlagged.C,v     demos.bash,v                  splitdepth.py,v
DensityFlagged.H,v     density-sml.jpg,v             vol2vol.C,v
Doxyfile,v             density.info.in,v             vol2vol_cmd.ggo.in,v
Eigs.C,v               eigs2xyz.py,v                 vol_iv.C,v
Eigs.H,v               endian.C,v                    vol_iv_cmd.ggo.in,v
HEADER-files.html,v    getgroups-ol92.bash,v         volhdr_edit.C,v
HEADER.html,v          histogram.C,v                 volhdr_edit_cmd.ggo.in,v
INSTALL,v              histogram_cmd.ggo.in,v        volinfo.C,v
InventorUtilities.C,v  hsva.cpt,v                    volinfo_cmd.ggo.in,v
InventorUtilities.H,v  is_equal.c,v                  volmakecmap.C,v
LICENSE.GPL,v          kdsPmagL.H,v                  volmakecmap_cmd.ggo.in,v
Makefile,v             makeCDF.C,v                   xyz_iv.C,v
Makefile.endian,v      makemovie.bash,v              xyz_iv_cmd.ggo.in,v
README.txt,v           one.bash,v                    xyzdensity.C,v
SiteSigma.C,v          regression.bash,v             xyzdensity.help2man,v
SiteSigma.H,v          render.C,v                    xyzdensity_cmd.ggo.in,v
TODO,v                 render.in,v                   xyzvol_cmp.C,v
VERSION,v              render_cmd.ggo.in,v           xyzvol_cmp_cmd.ggo.in,v
VecAngle.C,v           rgba.cpt,v
VecAngle.H,v           rosenbaum-ams-stripped.dat,v

That should give you some sense of how the svn tree is built based off of the rcs formated ,v files.

I am working on understanding how to use svn's ignore tag so that svg status skips things like python byte compiled files (*.pyc). This does not work the same way that .cvsignore works. Here is a quick example.

$ svn status
?      version.pyc
?      verbosity.pyc
?      BitVector-1.2
?      dist
?      html
?      MANIFEST
?      ais.pyc
M      ais.py



$ cd && svn propset svn:ignore "*.pyc" .
property 'svn:ignore' set on '.'



$ svn status
?      BitVector-1.2
?      dist
?      html
?      MANIFEST
 M     .
M      ais.py

That is great, but I also want to ignore the html subdirectory.

$ cd && svn propset svn:ignore "html" .
property 'svn:ignore' set on '.'



$ svn status
?      version.pyc
?      verbosity.pyc
?      BitVector-1.2
?      dist
?      MANIFEST
?      ais.pyc
 M     .
M      ais.py

Whoops. *.pyc came back to life! Instead, I discovered that propedit will bring up my text editor to do the job and that new lines between each entry works.

  svn propedit svn:ignore .

I then put in this

html
*.pyc
MANIFEST
dist
BitVector*

And now my svn status is much nicer.

$ svn status
 M     .
M      ais.py

Python Testing Tools Taxonomy

Bitten looks especially interesting.

  Bitten is a Python-based framework for collecting various software
  metrics via continuous integration. It builds on Trac to provide an
  integrated web-based user interface.

I just put together a quick and dirty fink info package for epydoc 3.0alpha. I want to give it a try on a new python module I am working on and pydoc really is not up to what I would like to do.

epydoc-py.info

Doh... why doesn't fink validate know about the MIT license?

I am finally converting my CVS repository that I used throughout my time working on my thesis into a subversion (svn) repository. Here is what I had to do. The first step was to tar up the CVS repository on my server. Names are ever evolving, so I will just put them in for historical kicks.

ssh my.server.com
#
cd projects/xcore/
ls -l Master
total 0
drwxr-xr-x  38 schwehr  schwehr  1292  4 Nov  2002 CVSROOT
drwxr-xr-x   8 schwehr  schwehr   272 21 Mar 09:34 dotfiles
drwxr-xr-x   5 schwehr  schwehr   170 12 Feb 10:58 xcore
#
tar cf Master.tar Master
bzip2 -9 Master.tar
mv Master.tar.bz ~/
exit
#
cd ~/Desktop/
mkdir ForSVN
cd ForSVN
scp my.server.com:Master.tar.bz2 .
tar xf Master.tar
#
### Rearrange things to be happier inside of Master
#
ls -l Master Master/src Master/doc
drwxr-xr-x   38 schwehr  staff        1292 Nov  4  2002 CVSROOT
drwxr-xr-x    7 schwehr  staff         238 Mar 21 13:15 doc
drwxr-xr-x    8 schwehr  staff         272 Mar 21 08:42 dotfiles
drwxr-xr-x    8 schwehr  staff         272 Mar 21 13:13 src
drwxr-xr-x    5 schwehr  staff         170 Feb 12 13:58 thesis
#
Master/doc:
-r--r--r--    1 schwehr  staff      448433 Apr 12  2005 kurt-2004.txt,v
-r--r--r--    1 schwehr  staff      273650 Feb 20 12:50 kurt-2005.txt,v
-r--r--r--    1 schwehr  staff       53270 Mar 21 10:45 kurt-2006.txt,v
drwxr-xr-x    8 schwehr  staff         272 Mar 21 10:44 segy-py
-r--r--r--    1 schwehr  staff       10016 Mar 21 10:45 todo.txt,v
#
Master/src:
drwxr-xr-x   51 schwehr  staff        1734 Mar 21 11:26 Terrain
drwxr-xr-x   92 schwehr  staff        3128 Mar 21 11:26 density
drwxr-xr-x    6 schwehr  staff         204 Mar 21 11:26 mypy
drwxr-xr-x   27 schwehr  staff         918 Mar 21 11:26 pmag-kds-py
drwxr-xr-x   47 schwehr  staff        1598 Mar 21 11:26 segy-py
drwxr-xr-x    6 schwehr  staff         204 Mar 21 11:26 xsonar
# 
cd Master/src && mv segy-py seismic-py
cd Master && rm -rf CVSROOT # This is why I work on a copy
mv Master schwehr
mkdir svnrepo

Now do the conversion, but the first time I tried, it failed like this:

$ cvs2svn -s svnrepo/schwehr schwehr
yada yada
===== pass 2 =====
Checking for blocked exclusions...
Checking for forced tags with commits...
Checking for tag/branch mismatches...
ERROR: The following symbols are tags in some files and branches in others.
Use --force-tag, --force-branch and/or --exclude to resolve the symbols.
    'start' is a tag in 351 files, a branch in 4 files and has commits in 4 files.

So I then tried

cvs2svn --force-tag=start -s svnrepo/schwehr schwehr
...
===== pass 2 =====
Checking for blocked exclusions...
Checking for forced tags with commits...
ERROR: The following branches cannot be forced to be tags because they have commits:
    'start'

So I made them all branches. I hope this does not give me trouble.

cvs2svn --force-branch=start -s svnrepo/schwehr schwehr
tic tock...
cvs2svn Statistics:
===============---
Total CVS Files:              2365
Total CVS Revisions:          8885
Total Unique Tags:              13
Total Unique Branches:           9
CVS Repos Size in KB:       110389
Total SVN Commits:            5591
First Revision Date:    Sat Jul  6 13:05:09 2002
Last Revision Date:     Tue Mar 21 11:27:42 2006
===============---
Timings:
===============---
pass 1:    52 seconds
pass 2:     2 seconds
pass 3:     0 seconds
pass 4:     3 seconds
pass 5:    20 seconds
pass 6:     0 seconds
pass 7:     0 seconds
pass 8:   964 seconds
total:   1043 seconds

And finally, to import my stuff into the new svn tree... first to see what dump-only does. This file is pretty large. My 30MB CVS repository jumped to 180MB

cvs2svn --dump-only --force-branch=start --dumpfile schwehr.svndump svnrepo/schwehr

So maybe build the dump the way roland did:

svnadmin dump schwehr --deltas > schwehr.svndump.deltas

And testing out the repository, I am doing a local checkout.

cd
mkdir test
cd test
svn checkout file://localhost/home/schwehr/svnrepo/schwehr
ls -la schwehr/
  total 12
  drwxrwxr-x   7 schwehr schwehr 4096 Mar 21 14:55 .svn/
  drwxrwxr-x  13 schwehr schwehr 4096 Mar 21 14:54 branches/
  drwxrwxr-x  15 schwehr schwehr 4096 Mar 21 14:55 tags/
  drwxrwxr-x   8 schwehr schwehr 4096 Mar 21 14:54 trunk/

More later when this is spliced into the server.

Mapping Human Activity in the Marine Environment [slashgeo]

  This report summarizes the results of the workshop (convened by the
  National Marine Protected Areas (MPA) Center Science Institute) that
  brought together 20 social scientists, geographers, and GIS
  specialists from government agencies, universities, and
  non-governmental organizations with expertise in characterizing
  human use patterns in the marine environment with the aid of GIS
  tools. The goal of the workshop was to develop general design
  criteria for a practical participatory method or a suite of methods
  to collect spatial data on human use patterns to inform local and
  regional MPA planning processes. 

Bummer... this is mostly abstracts. Maybe there is good stuff in the publications section of the US Marine Protected Areas web site.

Other news... SAIC Celebrates Opening of Advanced Radar Research and Development and Production Center in Florida [gisuser.com]

mapbender

  Mapbender is all about maps! Have a look at this map
  application. You can zoom in, out, pan, click and query, turn layers
  on and off or add more maps. You can also add new points to the map
  with the flag button, all Mapbender operators are invited to do
  so. There are many different possibilities to make use of maps, see
  some more examples in the Mapbender Gallery. Please observe that
  this software only shows and manages maps. The maps themselves come
  from many different Map Services all over the world. 



  Mapbender is the software and portal site for geodata management of
  OGC OWS architectures. The software provides web technology for
  managing spatial data services implemented in PHP, JavaScript and
  XML. It provides a data model and interfaces for displaying,
  navigating and querying OGC compliant map services. The Mapbender
  framework furthermore provides authentication and authorization
  services, OWS proxy functionality, management interfaces for user,
  group and service administration in WebGIS projects.  

Not in fink, but it looks interesting.

I just went and read the GNU make info page on ".PHONY" targets. I have read this many times, but I guess I never really internalized what it means I should do. Phony rules are those that do not depend on files being created. The classic is the clean rule, but I think test will be more informative. Normally a test rule might look something like this:

test:
	cd test && make test

or this (depending on your style preference)

test:
	make -f test/Makefile test

If the test directory exists, make will see that directory and think that everything for test is already done. make will say:

make: `test' is up to date.

Clearly, that is not what needs to happen. Make should go run the test! By saying that test is a phony target, make will ignore the test directory and happily run the commands for that target. Here is a dummy test target.

.PHONY: test
test:
	cd test && pwd

Which when run will work just fine:

mkdir test
make test
  cd test && pwd
  SOME_LONG_PATH/test

And finally, the info page says this speeds up the make process because make is no longer looking for those targets. This is not a huge speedup, but in large projects this time will add up as the build system gets larger and if building happens frequently.

Soon I will be trying to run multiple operating systems on top of Mac OSX. I bought a copy of Virtual PC. Virtual PC also gives me a legal copy of XP pro. I will also be trying Q [kju:] which is QEMU.

It does not look like VMWare works on the Mac.

It is nice to see that there are premade images for Virtual PC of things like Ubuntu (The Breezy Badger) - 5.1.0; Linux i386 disk image for Virtual PC

Others include:

• bochs
• PearPC
• wine/darwine
• xen
• GuestPC

COSMIC is the old NASA code release program. You got to spend a lot of money on software. I doubt that this program took in more money than it spent. Plus I still don't see VEVI in their list of programs and I know VEVI went into COSMIC.

GCN just posted about http://opensource.arc.nasa.gov. Why is there yet another open source license? And why is not more NASA developed software in there? So much code that is written at NASA rots or dies as it sits around doing nothing. I would love to see any of the projects I worked on at Ames go open source... even in those that probably don't work. For example, we worked really hard on a bunch of VxWorks and Linux software for the Marsokhod rover. Much of it has been superseded as new development teams take over after my group mostly left. That's fine, but that old work is probably just gone. It would be great to allow people to view the source code as the read the old journal articles!

Mercury Computer pitches game processor for military use [Government Computer News]

...
  The Chelmsford, Mass., company is confident that the Cell Broadband
  Engine, as it is properly known, cannot only replace existing
  signal-processing technologies, but also introduce real-time radar,
  sonar and signals intelligence processing into the battlefield
  itself. 



  The gaming machine processor "is a glimpse of the future," enthused
  Craig Lund, Mercury chief technology officer. "It's a disruptive
  change."
...

I am not sure why this is a "disruptive change" as general CPU's are becoming more and more able to do these kinds of DSP tasks. We don't need the TI DSP chips as much anymore. Many of us have been thinking along these lines (e.g. Alex) since at least 1999, but not for specifically military applications. The Cell architecture is going to take a couple years to come into its stride as the compiler and other tools get better.

I'm trying to build a flexible parser for AIS. There are a number of issues with the AIS specification that make it hard to parse. AIS is definitely NOT something that a normal grammar parser can handle (e.g. LALR and friends). One example is the region section of the position messages which are 1, 2, and 3. One of the fields is a region specific 4 byte value. How is the parser to know what region it is in? And then there is the "bit stuffing" mode that I have not gotten to yet where the message must have extra bits "stuffed" in there. I have yet to figure out why this is, but I am suspecting that certain bit patterns will be misconstrued as the training dataset at the beginning of each AIS message.

I've decided to go the route of a large setup of lookup tables that embed parse functions that know what to do with a chunk of bits. This is going to get challenging when the fields have interdependencies, but I will just have to refactor at that point.

To make my life safer, I am applying asserts to many of the functions to validate all the input and make python a bit more type safe. I know, this is not very pythonic, but I like to catch errors early. This is currently my function to build one table entry.

fieldTypes=['int']
def makeEntry(offset, fieldname, numbits, description='N/A', type='int', parseFunction=parseInt, units='N/A'):
    """Create the lookup table entry



    offset - index of the first bit
    fieldname - what is this field called.  e.g. FIX
    numbits - number of bits for this field.  FIX: what if variable??  Scary.
    type - dict, int, float??, string, ?? what else.  Standard types that are handled.
        dict must have a parseFunction!
    parseFunction - function that handles the contents of these bits
    units - the units after being converted... e.g. degrees/minute
    """
    assert(offset<1000) # FIX: what is the max offset?
    assert(offset>=0)
    assert(numbits>0)
    assert(numbits<1000) # FIX: what is the max message size?
    assert(fieldTypes.__contains__(type))  # Make sure this is allowed
    assert(parseFunction != None)
    import types
    assert(isinstance(parseInt,types.FunctionType))
    # Build the entry now that the world is safe
    entry = {'fieldname':fieldname,
	     'startOffset': offset,
	     'endOffset': offset+numbits-1,
	     'numbits': numbits,
	     'description':description,
	     'type': type,
	     'parseFunction':parseFunction}
    return(entry,offset+numbits)

The first few asserts just check the range of the offset and numbits. These need to be constrained to the maximum sizes of AIS packets. For the 'type' value, this has to be something the code knows about. I want to be able to expose a list of known types to people using this library - the 'fieldTypes' variable.

The trickiest bit is making sure parseFunction is an executable function. I would rather fail when building the table than go a long time before some rare message hits a bad value for the parseFunction.

And units... I have got to track these. It is critical to know what units a value has. There are too many running around in the marine world. SI, CGS, and nautical. What if someone assumed a depth was in fathoms and it really is feet? That would be 'bad'.

To build up a message dictionary, I do something like this:

    msg={}
    offset=0
    name='MessageID'; entry,offset=makeEntry(offset,name,6); msg[name]=entry
    name='RepeatIndicator'; entry,offset=makeEntry(offset,name,2); msg[name]=entry
    name='UserID'; entry,offset=makeEntry(offset,name,30,'MMSI ship id'); msg[name]=entry
    name='NavStatus'; entry,offset=makeEntry(offset,name,4); msg[name]=entry

I am now in a waiting pattern for a couple fink packages to make it into the CVS tree. I have been enjoying some python coding using the BitVector module.

C++ Command Pattern for Network Operations reminds me that I should pick up my Design Patterns book again and start reading.

From QuickDraw to Quartz 2D reminds me that I am not really a Mac programmer. I am a software engineer, but I try to avoid too much Mac specific stuff.

Types and Objects in Python - Looking at the BitVector class shows that it is not too hard to create new solid types.

I still do not feel comfortable with this kind of thing:

ipython
In [1]: l = ['1', 'two', '3']
In [2]: l = ["prepend" + s for s in l] 
In [3]: print l
['prepend1', 'prependtwo', 'prepend3']

And this is even stranger.

In [4]: l = ['1', 'two', '3']
In [6]: map("prefix%s".__mod__,l)
Out[6]: ['prefix1', 'prefixtwo', 'prefix3']

Now I'm working on cppunit for fink because gnuradio-core requires a newer version than is in fink. Also ran into trouble with fftw3. The tar ball for fftw 3.0.1 disappeared from the ftp server. Doh! Instead, I just used the binary version with a "sudo apt-get install fftw3 fftw3-shlibs"

I seem to have a case of disappearing xterm windows. Is it just after a sleep? I open a xterm from the Applications menu, it then flashes on the screen and is gone. If I open one a few seconds later, it opens like normal. What gives?

Working on packaging gnu radio for fink. This one may take a while. Shared libs and a whole bunch of separate packages.

http://www.cocoalab.com/developer.htm - How to Become An Xcoder Book - free 60 page pdf book.

I have never really gotten into the whole cocoa programming thing. Maybe I should read this.

http://www.gnu.org/software/gnuradio/

December 9, 2005 - GNU Radio 2.6 release



GNU Radio 2.6 includes sending and receiving data at up to 1Mbit/sec
using GMSK. There's also support for the new USRP Flex 400 transceiver
daughterboards. Handling of all daughterboards has been unified, so
most apps will work without modification on any of them. For those of
you interested in wireless networking, see especially
gnuradio-examples/python/gmsk2. We've got a framework in place to
allow you to build and experiment with your own MACs. 2.6 includes the
"null MAC". A carrier sense MAC is the next logical step. For
additional details see http://comsec.com/wiki?GnuRadio2.X.

Will it work on the Mac? It appears so: http://staff.washington.edu/~jon/gr-osx/gr-osx.html

http://www.gisuser.com/content/view/8430/

 Database-driven Additional Military Layer Product Editor from CARIS
Written by CARIS 



CARIS announces the release of the Additional Military Layer (AML)
Product Editor, the latest addition to the CARIS Hydrographic
Production Database (HPD) product suite.Fredericton, NB, Canada -
March 16, 2006 - CARIS announces the release of the Additional
Military Layer (AML) Product Editor, the latest addition to the CARIS
Hydrographic Production Database (HPD) product suite.



The release of AML Product Editor expands upon the existing HPD
product suite, which already includes the HPD Source Editor, S-57 ENC
Product Editor and Paper Chart Editor.

AIS Range Performance Being Monitored by R&D



An experimental Automatic Identification System (AIS) reception
network has been established to analyze range performance and improve
the ability to monitor vessel movement.  The network, consisting of 20
reception sites at locations from Massachusetts to Hawaii, is used to
study potential AIS surveillance performance.  The live vessel data
obtained by the network are distributed to a number of experimental
and operational government information systems, such as the USCG
Common Operational Picture.  The AIS signals are generally thought to
have a short operational range - only a little greater than the
visible horizon.  Experimentation with receiver design improvements,
gain antennas, and attention to radio noise reduction has shown that
AIS surveillance systems can frequently receive signals at ranges 2 to
10 times the visible horizon.  Site-by-site analysis of range
performance for all the reception sites connected to the RDC network
is now available on an experimental web page
https://ra2.rdc.uscg.gov/NAISprototype/AISdata.htm.  The analysis
should be of interest to VHF communications system planners, such as
those designing high performance AIS facilities and surveillance
networks.



Project: Automatic Identification System Research and Development (2410.5)

Vessel Tracking Project (VTP) Moves Toward Transition



The USCG and the U.S. Navy have joined forces in a cooperative
partnership, funded with more than $31M of Navy money, that will
provide the capability for extracting information from multiple
security sources and provide the capability to fuse multi-intelligence
data sources.  The Vessel Tracking Project (VTP) will be deployed as
an operational prototype system to support Maritime Domain Awareness.
The cornerstone of this project is the Common Distributed Virtual
Database with Information Extraction (CDVD/IE).  The CDVD/IE allows a
fusion of data with high levels of automation to create highly
detailed vessel tracks.  The R&D Program, the primary CG liaison to
the Naval Research Lab (NRL), has exerted significant influence over
this project so that it will provide operational value and near-term
return on investment for the CG.  The project's first operational
capability will be deployed at the Maritime Intelligence Fusion Center
(MIFC) Atlantic in July 2006, and additional installations are planned
for Maritime Intelligence Fusion Center (MIFC) Pacific and National
Maritime Intelligence Center (NMIC).  From a long-term transition
perspective, initial planning meetings have resulted in the
identification and chartering of a joint USCG-USN Transition Working
Group.



Project:        Department of the Navy USCG Vessel Tracking Project

http://www.seacoastonline.com/news/03162006/news/92812.htm

  NEWINGTON - Port officials and the U.S. Army Corps of Engineers are
  entering the next phase of planning the expansion of the 800-foot
  turning basin in the Piscataqua River. 



  Director of Ports and Harbors Geno Marconi said a new study will
  determine what type of sediment will need to be removed, the best
  way to remove it and where it will be sent in order to widen the
  basin. 



  Expanding the basin, located in the federal navigation channel
  between the shores of Eliot, Maine, and Newington, has been
  discussed for several years.
...

http://www.macresearch.org/survey_the_transition_to_intel_and_its_impact_on_scientific_computing

Here are some of my responses to that request.

Please merge your GNU Make tweaks back into the mainline make tree!!
It is really frustrating that the standard GNU Make does not work for
the Fink project.



Would be good if Apple would contribute more to the DarwinPorts and
Fink projects.  Continual support and improvement of X11 is critical.



The mac package management/receipt system still needs some help.



The firewire camera support is still too complicated.  How about an
option to work like the old Linux BTB848 drivers where programs can
read images from /dev/camera.format (e.g. ppm or pgm).



As a scientist, I just don't have time to struggle through learning
Cocoa or Carbon enough to get at some of these APIs.



Please get python to to 2.4.x!



The lake of case sensitivity is driving me nuts with the Mac.  The
macosx libJPEG is messing with libjpeg.  That kind of stuff is really
painful.



More, more, more 64 bit support.



I would rather stick with PowerPC than go to Intel.



More CPUs, PCIExpress slots, and RAM please.  I want a 16-way PowerPC
box with 128GB of RAM and 4 PCI-Express 16-lane graphics cards that
all are stereo capable.



How about a Mac OS-X based data-logger/embedded machine.  Okay, I know
that is crazy and I would rather use Linux where I can control
everything, but would be interesting.



Internal RAID arrays would be "huge."



Why not just put the server functionality in every copy of Mac OSX?

I would also like to see ports to the Mac (both x86 and PPC) of ESRI tools, Caris, and SGI Performer. Not for me, but for others so that there is a chance of going all Mac for operations (e.g. on-board ships and spacecraft mission control centers)

The World of Smartboards, Sympodiums is about to change by Jefferson Y. Han

Thanks to Roi for pointing this out.

Avi tweaked BitVector a little bit to allow the binary or operator to work on different sized BitVectors. This is great as it make my function to set grow size of a BitVector that much easier.

In [1]: from BitVector import BitVector
In [2]: print BitVector(intVal=1) ^ BitVector(intVal=2)
11
In [3]: print BitVector(intVal=1) ^ BitVector(bitlist=[0,0,0,0,0,0,0,0])
00000001
In [4]: print BitVector(intVal=1) ^ BitVector(bitlist=[0]*8)
00000001

So the new function is just two lines...

def bvSize(bv,size=8):
    return (bv ^ BitVector(bitlist=[0]*size))

Which works like this:

In [6]: print bvSize(BitVector(intVal=123),32)
00000000000000000000000001111011

Abstracts are due in 5 days...

• IEEE Visualization 2006 - http://vis.computer.org/vis2006/
• IEEE InfoVis 2006 http://conferences.computer.org/infovis/infovis2006/
• IEEE VAST 2006 http://conferences.computer.org/vast/vast2006/

TimeDrive: The fourth dimension comes to Google Earth

I just added BitVector 1.2 to fink 10.4-transitional/unstable. This is a nice package for bit twiddling written by Avi Kak at Purdue.

This package makes working with bit level data much easier than it ever was in C or C++. STL in C++ has trouble with bit vectors and the C bit operators are not the most fun. BitVector has the muscle to make the job much more understandable.

python
from BitVector import BitVector
bv = BitVector(bitlist = [ 0,0,1,0]); print bv
  0010
print bv[-2]
  1
b2=BitVector(intVal=2);print b2
  10

Gigabyte GA-8N-SLI Quad Royal Motherboard has 4 PCI Express x16 slots. Too bad this is not the same motherboard that has 4 opteron sockets. Just think about 8 Opteron Cores with 128 GB of ram and 4 Graphics Cards. Now combine that with a math library that is able to push some operations to the graphics cards. That is some serious crunching power in a small form factor without going with crazy hardware. What kind of powersuppy and cooling system would be needed? Perhaps just hook it straight into the buildings AC system.

This is VERY exciting! After wanting a swig python API for OpenInventor for 9 years, I just ran one for the first time today. I now have fink info files that work for 10.4-transitional.

• swig1325.info
• pivy-py.info

AWESOME!!!

#!/usr/bin/env python
# Copyright (c) 2002, Tamer Fahmy <tamer@tammura.at>
import sys
from pivy.sogui import *
from pivy.coin import *
def main():
    myWindow = SoGui.init(sys.argv[0])
    if myWindow == None: sys.exit(1)
#
    root = SoSeparator()
    myCamera = SoPerspectiveCamera()
    myMaterial = SoMaterial()
    root.addChild(myCamera)
    root.addChild(SoDirectionalLight())
    myMaterial.diffuseColor = (1.0, 0.0, 0.0)   # Red
    root.addChild(myMaterial)
    root.addChild(SoCone())
#
    myRenderArea = SoGuiRenderArea(myWindow)
#
    myCamera.viewAll(root, myRenderArea.getViewportRegion())
#
    myRenderArea.setSceneGraph(root)
    myRenderArea.setTitle("Hello Cone")
    myRenderArea.show()
#
    SoGui.show(myWindow)  # Display main window
    SoGui.mainLoop()    # Main Inventor event loop
#
if __name__ == "__main__":
    main()

No, this is now World Tool Kit... this is worldKit - a flash based google maps like system that is released under the GPL license. Now quick as slick an interface as Google maps/local, but it looks like it works. I am curious about the RSS integration, but don't have time right now to dig further.

Note: I am still working on pivy and bitvector packages. For pivy, Tamer let me know that it will not work with swig 1.3.28, so I am packaging swig 1.3.25. Hopefully swig 1.3.29 will clear things up.

I heard this was going to be released soon and now I can finally talk about it.

        Google Mars

Then click on spacecraft. You will see where all the spacecraft have gone. Now where are Zareh's movies? If you go to the about page, you will eventually end up here:

        Flight Into Mariner Valley: The Movie

I got to try the HDTV sized version of the movie and it bought my laptop to its knees, but the smaller versions play perfectly Great movie!

I saw an article today about setting up buildbot to manage servers to build latest version of software for a project on one or more machines and decided to update my fink package that I started about a year ago. I upgraded it to version 0.7.2 and made it now depend on python 2.4, twisted2-web2, and zopeinterface-py24. I only did the zope thing because of the articles below. Still have not worked through them. Need to get my new mac server! I can't wait to have automated building and testing. Maybe I can even do Mac, Windows, and Linux all on the same machine using virtualization.

buildbot-py.info

The articles on setup are:

• Continuous integration with buildbot
• Running buildbot on various platforms

As usual, you will need to put the info file in /sw/fink/10.4-transitional/local/main/finkinfo/devel and download the source from the buildbot home page. The tar.gz goes in /sw/src/

The title says it all. SigGraph2006 is in Boston which is just down the road from here. Wow... it is already too late to submit papers.

This sounds like it has some pretty decent compute power... 4 socket 1U opteron system with 4 dual core chips: Supermicro stuns with four-socket, 1U Opteron

http://www.supermicro.com/Aplus/motherboard/Opteron/nForce/H8QC8+.cfm

That would make a nice little cluster with a bunch of these. Issues are just power and cooling.

I submitted a change request to the swig team asking for a fix so that swig honors the --program-suffix and --program-prefix configure options. Wow, what a quick turnaround!

William Fulton (wsfulton)
Date: 2006-03-10 22:58



Message:
user_id=242951



SWIG has a non-standard way of doing this, use the
--with-release-suffix option. For 1.3.29, I've removed this
options and fixed the standard ones, namely,
--program-prefix and --program-suffix

We have a VLinx device that is reading RS422 data and spitting out UDP broadcast packets. Yesterday, I got pretty confused trying to understand what it was doing by looking at the tcpdump output. Now that Roland has a given me his nice little SDL based application to read UDP traffic, the tcpdump stuff is not as important, but here is the tcpdump line I just came up with.

sudo tcpdump -i en0 -A -n src IP-ADDRESS-OF-VLINX and udp port 4003

• The "-i en0" reads my mac laptops built in ethernet port.
• "-A" dumps an ascii version of the package payload
• "-n" says just give the numeric port and ip numbers
• "src IP-ADDRESS-OF-VLINX" only receive packets from that address
• "and" logical glue to the next statement
• "udp port 4003" says we only want to see that port.

This gives results like this...

15:27:54.796096 IP 10.48.2.199.4003 > 255.255.255.255.4003: UDP, length: 14
E..*4.@...9.
0
        ............s.!AIVDO,1,1,,,1rks 27..
15:27:54.800100 IP 10.48.2.199.4003 > 255.255.255.255.4003: UDP, length: 14
E..*4.@...9.
0
        ..............018lEgP?w<tSF0rks .G..
15:27:54.803785 IP 10.48.2.199.4003 > 255.255.255.255.4003: UDP, length: 14
E..*4.@...9.
0
        ..............l4Q@>4?wp1d00,rks d.f\
15:27:54.805363 IP 10.48.2.199.4003 > 255.255.255.255.4003: UDP, length: 6
E.."4.@...9.
0............_)0*78

Which is not as nice as

@GZTIM,185.807016
!AIVDO,1,1,,,1018lEgP?w4?wp1l00,0*70
@GZTIM,186.858809
!AIVDO,1,1,,,1018lEgP?w4?wp1l00,0*70
@GZTIM,187.907003
!AIVDO,1,1,,,1018lEgP?w4?wp1d00,0*78
@GZTIM,188.955972
!AIVDO,1,1,,,1018lEgP?w4?wp1d00,0*78
@GZTIM,190.007628
!AIVDO,1,1,,,1018lEgP?w4?wp1l00,0*70
@GZTIM,191.055809
!AIVDO,1,1,,,1018lEgP?w4?wp1h00,0*74

http://www.geonames.org/ - does it an a web api or a way to get all of the database?

Portland Maine map exihibit

http://www.georss.org/ - Geographically Encoded Objects for RSS feeds

Here is a nice example python module that illustrates 4 imporant concepts.

1. Using doctest is really easy. Lines 15-18 and 32-38 define some examples that doctest will validate. The python code starts with >>> and is followed by the expected output from that command. At the end of the program (lines 59-60), doctest is called. It will scan through the whole module and validate every doctest formatted example.
2. Using unittest is a little bit more work, Lines 48-55 are a basic unittest setup. This can be done much fancier with setup and teardown methods before and after groups of tests and much more.
3. On a different note, this module handles NMEA 0183 checksum calculation and validation for AIS and GPS strings. Not very exciting, but it provides an example and shows how the string is OR'ed (^ in python) together.
4. Finally, I used "pydoc -w nmeaChecksum.html" to produce the nmeaChecksum web documentation. I still need to give epydoc a try, as it looks more like doxygen. Pydoc is not all that powerful.

 #!/usr/bin/env python
 """
 Background info on NMEA strings can be had from here:
     http://gpsd.berlios.de/NMEA.txt
 """
 __version__ = '$Revision: 1.1 $'.split()[1] # See man ident
 __date__ = '$Date: 2006/03/10 10:11:40 $'.split()[1]
 __author__ = 'Kurt Schwehr'
 
 def nmeaChecksumStr(data):
     """ Take a NMEA 0183 string and compute the checksum 
 
     Checksum is calculated by xor'ing everything between ? or ! and the *
 
     >>> nmeaChecksumStr("!AIVDM,1,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0*09")
     '09'
     >>> nmeaChecksumStr("AIVDM,1,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0")
     '09'
     """
     
     if data[0]=='!' or data[0]=='?': data = data[1:]
     if data[-3]=='*': data = data[:-3]
     sum=0
     for c in data: sum = sum ^ ord(c)
     sumHex = "%x" % sum
     if len(sumHex)==1: sumHex = '0'+sumHex
     return sumHex
 
 def isNmeaChecksumValid(nmeaStr):
     """Return True if the string checks out with the checksum
 
     >>> isNmeaChecksumValid("!AIVDM,1,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0*09")
     True
 
     Corrupted:
 
     >>> isNmeaChecksumValid("!AIVDM,11,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0*09")
     False
     """
     if nmeaStr[-3]!='*': False  # Bad string without proper checksum
     checksum=nmeaStr[-2:]
     if checksum==nmeaChecksumStr(nmeaStr): return True
     return False
 
 ######################################################################
 # Test code
 ######################################################################
 import unittest
 class TestChecksum(unittest.TestCase):
     """Are we computing the checksome correctly?"""
     def testSimple(self):
 	self.assertEquals('09',nmeaChecksumStr("!AIVDM,1,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0*09"))
 	self.failUnless(isNmeaChecksumValid("!AIVDM,1,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0*09"))
 	# Corrupted
 	self.failUnless(not isNmeaChecksumValid("!AIVDM,11,1,,B,35MsUdPOh8JwI:0HUwquiIFH21>i,0*09"))
 
 if __name__ == '__main__':
     unittest.main()
     import doctest
     doctest.testmod()    

This one stumped me way to long. I figured that it has to be somewhere in python. This is one of those things that happens through casting in c.

ipython



In [1]: chr(97) 
Out[1]: 'a'



In [2]: ord('a')
Out[2]: 97

http://www.python.org/doc/lib/built-in-funcs.html

This looks very interesting on the 3D volume rendering/volex front. Hopefully I will get pivy going soon and maybe can work on adding voleon support to it. That would be a great combination. http://cheeseshop.python.org/pypi/Pyvox/0.72

  Pyvox (formerly known as BBLimage) is a set of software tools for
  medical image processing, particularly skull stripping and
  segmentation of MR brain images; tools to support other applications
  may be added later. These tools are intended to support researchers
  who need to prototype new image analysis algorithms or to develop
  automated image analysis tools for specific image analysis
  applications. The sequence of processing operations is specified
  through the scripting language Python, which can be used
  interactively or in command files; the core image processing
  algorithms are written in C for efficient processing of volume
  images.

In cleaning house, I have created gmtplus 3.4.2 which now builds with gmt 4.1.1 on Mac OSX 10.4.5. This code is suffering serious bit rot. uwmtogmtplus no longer compiles and I do not personally know how to use it or even test it. All I can say is that it compiles and that Steve Cande has given me some sample data to go along with the code. This will be my last attempt to do anything with the code, so I offer it up to the community. Unless someone puts in some time testing and debugging gmtplus this is the end of the road for it. I hear that a lot of the functionality in gmtplus is now in gmt, so hopefully the few of you that count on gmtplus can migrate to that.

I will be switching hosting services soon and will be loosing my unlimited bandwidth/storage limits, so if you want it, grab this stuff right away:

http://schwehr.org/software/gmtplus/

We have some progress on the shore based AIS station:

• Andy M. has the RM808 powered up on the bench and it seems happy.
• I got an mmsi identifier from http://www.boatus.com/mmsi/, which is free if you do it online.

The RM808 is one heck of a beefy unit!

I am starting to have strange flashbacks to IRIX/IRIX64/-old-ld/SGI death throws. I got this on OSX 10.4.5. I was trying to compile everything with gcc 3.3 since I am linking against all kinds of junk in fink 10.4-transitional that is c++ and compiled with gcc 3.3. I forgot that I need to do the same in a couple of external packages. Maybe that is why fprintf is not defined. Nuts! That should never happen for a straight C call. Bugger, bugger, bugger.

ld: warning multiple definitions of symbol _dct64
/sw/lib/libSDL_sound.dylib(libmpglib_la-dct64_i386.lo) definition of _dct64
/sw/lib/libmp3lame.dylib(dct64_i386.o) definition of _dct64
ld: Undefined symbols:
_fprintf$LDBLStub
_snprintf$LDBLStub
_vfprintf$LDBLStub
_vsnprintf$LDBLStub
_printf$LDBLStub
_sscanf$LDBLStub
make[2]: *** [GeoZui4D] Error 1

I even got desparate and added -lc to my link line which is now:

g++-3.3 -g3 -DDEV -Wall -Wno-deprecated -o my-app-please-work
10-billion-billion-object-files \
-L../external-libs/lib
-L../external-libs/lib/Tktable2.9 -L/usr/X11R6/lib -L./lib.Darwin
-L/sw/lib -lglut -lGLU -lXmu -lXext -lX11 -ljpeg -ltiff -lsmpeg
-lSDL_mixer -lSDL -lSDL_sound -lxml2 -lnetcdf_c++ -lnetcdf -lccom
-lgdal -lSDL_net -ltogl -lTktable2.9 -lavformat -lavcodec -lavutil
-ltk -ltcl -lvorbis -lz -logg -lmp3lame -lvorbisenc -lc -bind_at_load
-lSDLmain -lGL -framework Cocoa

I am currently working up to having a python/OpenInventor API working on my mac. I can't wait. This is something that Alex, Laurent, and I have wanted since early 1997.

Pivy needs swig >= 1.3.25, so I have created a local fink package and called it swig1328, but want to have this thing named correctly from the start. Along with the package name being different, I have the swig program named swig1328. What would be a better name/solution?

For a python package, how do I make setup.py use gcc3.3 when it compiles swig generated c++ code?

The reason I'm doing this is that swig 1.3.20 is drastically different than swig 1.3.28 and I hear that many programs can not handle the newer swig.

So close!

More fink updates. I am only committing changes to the 10.4-transitional tree right now. If you try any of my fink packages PLEASE SEND FEEDBACK! Either good or bad, I want to know! I have updated proj to 4.4.9 which just adds some geos projection stuff. I also upgraded pyproj to 1.7.

I took a look at wxglade packaging and realize that I need to learn more about the python distribution tools. All of my packages were really simple with everything at a top level. wxpython is a little bit more complicated.

Now I need to get the latest swig together so I can give pivy a try. If I remember right, swig is now radically different then at 1.3.20. I will be looking at swig-1.3.28.

Note to self: It would be really nice to have a blog feature that puts an expiration on some parts of the text. For example, I want to post a link to my gmt test packages for fink, but I want that info to disappear in a few days as it will all end up in the fink cvs tree and I will nuke that web directory when that happens.

Yes, I am pretty nerdy. I admit it. My work world is dominated by bash, emacs, and other Unix like programs. I enjoy listening to audio programs (aka podcasts and books on tape) while on the bus to and from work. I keep looking for great content, but what I find is occasional good content. A Very Spatial Poscast (AVSP) is not bad and the IT conversations are sometimes pretty interesting, but the best that I have found to date (for me at least) is The Linux Show (TLS) which is now defunct. It was a little over the top on open source politics, but it did have some tech content. What do I really want? I want 20-30 minute segments that attack some topic and take me 0-60mps in nothing flat. Teach me something cool. How about how to do something cool with gdb, emacs, python, whatever. If you know something really well, share that howto with us. Say you really understand the new '|' construct in GNU Make. Then tell us all what it is there for and how to properly use it. Are there any gotchas? Most podcasts seem to be people telling me how great they are and how many people they know or know them. Whatever. I just want to learn something fun and/or powerful that makes my life better.

I should lead by example right? I have to some extent. I had one of my visualization classes video taped and it was/is streamed from somewhere at UCSD if you want to see. I just don't remember where. I also try to do that with this blog... giving myself and others some insight in a length that is somewhat digestible.

If you know something, pick up that microphone, make yourself a script, and make a quick and dirty audio recording and teach us something. BTW, I could care less what is in iTunes for pod casts. What an annoying way to deal with podcasts. I feel like one of those old crusty mainframe guys I used to see wandering the basements of NASA back in the 1980's.

Hey Apple, how about a podcast on how to setup launchd?

fink install dict

We have been looking into bug tracking systems. I have had a rough time with bugzilla and want to avoid it. We have tested both MantisBTS and trac Within about 5 minutes I was really impressed with trac. It has a nice wiki and integrates with our svn tree... wow! What a pleasure to use. Here is an image of me looking at the attached diff to a trac item that relates to my top priority for GeoZui3D. I need to get wxglade into fink.

I'm made headway on a number of fink packages. My hope is to get things I need for mapping and visualization up to the latest/greatest and tested in the next week or two. Here is where I'm at on Monday morning.

I have GMT 4.1.1 building it seems to be working. I sent it off to a number of GMT folks just now for testing. It is to the point where I can build MB-System 5.0.9 against it. To me that is a pretty good sign! I am having trouble with 10.3. I am note very motivated to work on 10.3 since I only have access to a very slow 5 year old iMac running 10.3 and I'm focusing for my research on using 10.4 as my base platform. (Yes, I will be running virtual PC too and am hoping that I can get Linux going there.) So here is my problem with GMT 4.1.1 in 10.3. Anyone who wants to offer a solution, please email me!

  grdraster.c: In function `main':
  grdraster.c:324: error: structure has no member named `xy_off'
  grdraster.c:325: error: structure has no member named `xy_off'
  grdraster.c:327: error: structure has no member named `xy_off'
  grdraster.c:327: error: structure has no member named `xy_off'
  ... repeats for a number of source files.

I am going to try to beef up mbsystem so that it has a -dev splitoff with the static libraries and headers, but I am not sure how that will go and if it will make it into the fink repositories.

On the Coin/SoQt front, Kyrah, Chris S., and I have things pretty much all working. There seems to be a bug with the SoQt 1.3.x code that prevents both static and shared library generation at the same time. The support from the SIM/Coin folks is always awesome!

I am going to try to go for pyproj and proj very soon. pyproj is way out of date at 5 versions back, so that is a high priority item for me. Projections are one of those things that anyone working in spacial sciences ends struggling with way too often. More and better tools is always a good thing.

Kyrah got me to back up and start with the configure problem and that looks like was the key to getting shared libs to build for SoQt from cvs using Mac OSX 10.4.5/fink 10.4-transitional. Here is how what I did that works for me with Coin 2.4.4 installed from 10.4-transitional. This requires you to be using bash as your shell. I remember no to long ago when this kind of syntax made no sense to me! This forces the use of GCC 3.3 for the compilers.

wget http://ftp.coin3d.org/coin/src/snapshots/SoQt-latest.tar.gz
tar xf SoQt-latest.tar.gz
cd SoQt
QTDIR=/sw CC=gcc-3.3 CXX=g++-3.3 ./configure --without-framework \
  --enable-darwin-x11 --with-qt=/sw --enable-shared
make



ls src/Inventor/Qt/.libs/*.dylib
  src/Inventor/Qt/.libs/libSoQt.20.3.1.dylib
  src/Inventor/Qt/.libs/libSoQt.20.dylib
  src/Inventor/Qt/.libs/libSoQt.dylib



otool -L src/Inventor/Qt/.libs/libSoQt.dylib 
  src/Inventor/Qt/.libs/libSoQt.dylib:
        /usr/local/lib/libSoQt.20.dylib (compatibility version 24.0.0, current version 24.1.0)
        /sw/lib/libqt-mt.3.dylib (compatibility version 3.3.0, current version 3.3.5)

The question is now, can I make both shared and static libraries at the same time? Looks like appending --enable-static turns off shared library generation right now.

pysight is for grabbing isight images from within python!!! I have got to try this out. Who is this guy who does http://livingcode.org. Sounds like he is doing quite a number of interesting projects.

I just put together a new release of density. Thanks to Chris and Kyrah for helping me figure out how to get it to work on 10.4. My issue with JPEG from before is because of case insensitivity in the linker. I really prefer case sensitive machines!! The trick is to use the new DYLD_LIBRARY_PATH variable and make sure that DYLD_LIBRARY_PATH is unset.

unset DYLD_LIBRARY_PATH
if [ -z "$DYLD_FALLBACK_LIBRARY_PATH" ]; then
    export DYLD_FALLBACK_LIBRARY_PATH=@FINK@/lib
else
    export DYLD_FALLBACK_LIBRARY_PATH=@FINK@/lib:$DYLD_LIBRARY_PATH
fi

You can find it at density.

Doh! Today has been a tough one. I have been trying to multitask my way through a large number of issues to get my development environment back to a happier state. It is a really good thing that I did not upgrade before I finished my PhD. Here are some of the battles.

GeoZui started using the new "|" rule dependency construct as a easy way to avoid always running configure in each subdir. There may be other ways around this, but this one was simple, so Roland went for it. I figured that I could just update the fink make to be 3.80 and all would be well. That was easy enough, but Max Horn informed me that Apple has made all kinds of tweaks to their gnu make and they did not merge them back into the gnu make tree. So some things will not build with the stock gnu make. BAD APPLE! Please, please, please integrate your changes in or stop using them!

The next is my trouble with Coin3d and clashing libjpeg and the ImageIO framework. I can't follow the advice I found via google to just fink remove libjpeg. That would probably nuke quite a bit! This may be an simage issue, but I am not totally sure. I miss the linux ldd command for programs. What is the equivalent for the Mac? I did make an info file for coin45 2.4.4 which looks close except for this clash problem.

Then cam SoQt 1.3. It seems to be a bit cranky. Something was wrong with the man page install code and it never seemed to generate a shared library. Doh.

Then in working with GeoZui, it looks like the wxglade generated code from Roland is now getting nuked on clean and fink does not have wxglade. No problemo... except that wxglade does not have a setup.py, so I need to create one and hopefully the project will take it.

Yikes. But tomorrow I get to take a short road trip, so that should be fun.

http://mapdex.org... searched on Portsmouth and got a whole bunch of broken viewers, but this one worked.

orthomap

This comes from ArcNews which had a big full page writeup on it talking about how is is done with ArcIMS and WMS.

I was trying to get all of my tools back in shape with Mac OSX 10.4 (using fink 10.4-transitional) and I ran into a problem. This occurs with my coin/voleon viewer, simpleview, in density. I just updated the fink package for coin yesterday so it compiles, but that whole running thing is eluding me.

./simpleview



dyld: Symbol not found: __cg_jpeg_resync_to_restart
  Referenced from: /System/Library/Frameworks/ApplicationServices.framework/Versions/A/Frameworks/ImageIO.framework/Versions/A/ImageIO
  Expected in: /sw/lib/libJPEG.dylib
./simpleview: line 15:  4734 Trace/BPT trap          $cmd

Any suggestions? I tried rebuilding libjpeg in fink. No luck. Does this error mean that ImageIO is trying to use libjpeg and is unhappy? Or is it the other way?

If you want to comment, I have made a pasteboard entry... http://paste.lisp.org/display/17370. BTW, I think paste is a pretty handy tool!