Wordify

This code can count to a million in words. It could probably be refactored so it was shorter and could count to any number. If you were patient enough.

The "-b" flag puts "and" in the places a Brit would put it. It's optional but preferred.
I have random stopping points to increase the modest entertainment value of it running.

Can you recursively get wordify up to the quintillions? How should the "-b" flag be handled?

Can you do this in French?

from string import ascii_lowercase as alphabet
from sys import argv

units = [""] + "one two three four five six seven eight nine".split()
teens =  "ten eleven twelve thirteen fourteen fifteen sixteen seventeen eighteen nineteen".split()
tenties = 2 * [""] + "twenty thirty forty fifty sixty seventy eighty ninety".split() 


def name(number,terminate = True,continued = False,brit=False):
    assert type(number) == int
    assert number > 0 and number < 1001
    if number == 1000: return "one thousand"
    hundreds,residual = divmod(number,100)
    tens,ones = divmod(residual,10)
    result = ""
    if hundreds:
        result += units[hundreds] + " hundred"
    if residual and hundreds:
        result += " "
    if brit and residual and (hundreds or continued):
        result += "and "
    if (tens == 1):
        result += teens[ones] 
        if terminate:
            result += "."
    else:
        if tens:
            result += tenties[tens]
            if ones: result += "-"
        if ones:
            result += units[ones]
        if terminate:
            result += "."
    return result

def wordify(number,brit=False):
    assert type(number) == int
    assert number > 0 and number < 1000000
    thousands,residual = divmod(number,1000)
    result = ""
    if thousands:
        result = name(thousands,False,brit=brit) + " thousand"
        if residual:
            result += " "
            result += name(residual,continued = True,brit=brit)
        else:
            result += "."
    else:
        result = name(number)
    return result

if __name__ == "__main__":

    from time import sleep
    from random import randint

    brit = "-b" in argv
    
    sleeper = 30
    for i in range(1,1000000):
        if i == sleeper:
            sleep(3)
            sleeper += randint(3000,90000)
        print wordify(i,brit=brit)

output:

one.
two.
three.
four.
five.
six.
seven.
eight.
nine.
ten.
eleven.
twelve.
...
nine hundred and ninety-nine thousand nine hundred and ninety-six.
nine hundred and ninety-nine thousand nine hundred and ninety-seven.
nine hundred and ninety-nine thousand nine hundred and ninety-eight.
nine hundred and ninety-nine thousand nine hundred and ninety-nine.

That's as far as it goes so far. Go fix it.

Up to 11

I managed to confuse myself for hours on #11 trying to be elegant. But I really like this solution:

grid = [[0]*23]*3 + [[int(x) for x in line.split()]+[0,0,0] for line in
'''08 02 22 97 38 15 00 40 00 75 04 05 07 78 52 12 50 77 91 08
49 49 99 40 17 81 18 57 60 87 17 40 98 43 69 48 04 56 62 00
81 49 31 73 55 79 14 29 93 71 40 67 53 88 30 03 49 13 36 65
52 70 95 23 04 60 11 42 69 24 68 56 01 32 56 71 37 02 36 91
22 31 16 71 51 67 63 89 41 92 36 54 22 40 40 28 66 33 13 80
24 47 32 60 99 03 45 02 44 75 33 53 78 36 84 20 35 17 12 50
32 98 81 28 64 23 67 10 26 38 40 67 59 54 70 66 18 38 64 70
67 26 20 68 02 62 12 20 95 63 94 39 63 08 40 91 66 49 94 21
24 55 58 05 66 73 99 26 97 17 78 78 96 83 14 88 34 89 63 72
21 36 23 09 75 00 76 44 20 45 35 14 00 61 33 97 34 31 33 95
78 17 53 28 22 75 31 67 15 94 03 80 04 62 16 14 09 53 56 92
16 39 05 42 96 35 31 47 55 58 88 24 00 17 54 24 36 29 85 57
86 56 00 48 35 71 89 07 05 44 44 37 44 60 21 58 51 54 17 58
19 80 81 68 05 94 47 69 28 73 92 13 86 52 17 77 04 89 55 40
04 52 08 83 97 35 99 16 07 97 57 32 16 26 26 79 33 27 98 66
88 36 68 87 57 62 20 72 03 46 33 67 46 55 12 32 63 93 53 69
04 42 16 73 38 25 39 11 24 94 72 18 08 46 29 32 40 62 76 36
20 69 36 41 72 30 23 88 34 62 99 69 82 67 59 85 74 04 36 16
20 73 35 29 78 31 90 01 74 31 49 71 48 86 81 16 23 57 05 54
01 70 54 71 83 51 54 69 16 92 33 48 61 43 52 01 89 19 67 48'''.split('\n')
] + [[0]*23]*3

import operator

print max([reduce(operator.mul, [grid[y+n*d[0]][x+n*d[1]] for n in (0,1,2,3)])
for x in xrange(0,20) for y in xrange(3,23)
for d in ((0,1),(1,0),(1,1),(-1,1))]) 

The halo of zeros is very clever and completely works around what made mine confusing, and despite everything, inelegant:

linelen = len(array[0])
for line in array:
    assert len(line) == linelen

numlines = len(array)
offset = 4

directions = {
              "row":(0,linelen-offset+1,numlines,(0,1)),
              "column":(0,linelen,numlines-offset+1,(1,0)),
              "slash":(offset-1,linelen-offset+1,numlines-offset+1,(-1,1)),
              "backslash":(0,linelen-offset+1,numlines-offset+1,(1,1))
             }

maxprod = None
valmax = 0

for direction in directions.keys():
    x0, xlen, yrg, delta = directions[direction]
    vec0 = [(x0 + i * delta[0] ,i * delta[1]) for i in range(offset)]
    for y in range(yrg):
        vec1 = [(item[0]+y,item[1]) for item in vec0]
        for x in range(xlen):
            vec = [(item[0],item[1]+x) for item in vec1]
            vals = [array[item[0]][item[1]] for item in vec]
            prod = reduce(mul,vals,1)
            if prod > maxprod:
                valmax = vals
                maxprod = prod

print valmax, maxprod

Puzzle 5

Profiting from my overwrought example 3

from bf7 import *
from sys import argv

if __name__ == "__main__":
    if len(argv) == 1:
        upto = 10
    else:
        upto = int(argv[1]) 
    allprimes = primesupto(upto)

    result = 1
    for prime in allprimes:
        factor = prime
        while factor < upto:
            result *= prime
            factor *= prime
    print result

For me it would have been less work to just write

print 2 * 2 * 2 * 2 * 3 * 3 * 5 * 7 * 11 * 13 * 17 * 19

Best solution, for those who can remember grade school:

def gcd(a, b):
 while(b != 0):
  a, b = b, a%b
 return a
 
def lcm(a,b):
 return a * b / gcd(a, b)
 
print reduce(lcm, range(2, 21))

Puzzle 4

What's the largest palindrome number that is the product of two three digit numbers.

This time I have one of the better solutions:

def nextpal(low,high,reversed=False):
    assert low < high

    if reversed:
        candidate = high
        final = low
        increment = -1
    else:
        candidate = low
        final = high
        increment = 1
    while candidate != final:
        if str(candidate) == str(candidate)[-1::-1]:
            yield candidate
        candidate += increment

def getfactor(low,high,target):
    for test in range(low,high):
        if not target % test:
            residual = target/test
            if residual in range(low,high):
                return test, residual
    return None

if __name__ == "__main__":
    from time import time
    start = time()
    pal = nextpal(10000,1000000, reversed=True)
    while True:
        target = pal.next()
        result = getfactor(100,1000,target)
        if result:
            print target
            print result
            break
    end = time()
    print end - start

Most started from the factors to see if they built palindromes. Much better to go the other way; far fewer tests (the 94th palindrome succeeded, so less than 9400 calculations. The cast to string for the palindrome test is easy, but is it efficient?

The most interesting items on the discussion were the Pytho one-liner:

max(map(lambda s: bool(s) and max(s) or 0, [(filter(lambda n: n == int(str(n)[::-1]), map(lambda n: i*n, range(999, 99, -1)))) for i in range(999, 99, -1)]))

and the pen-and-paper solution

You can also do this with pen and paper. You have a number:

(100a + 10b + c)(100d + 10e + f)

Which is a palindrone. This factors out to:

100cd + 10ce + cf +
1000bd + 100be + 10bf +
10000ad + 1000ae + 100af

Assuming the first digit is 9, then cf must be equal to nine.
Also, both a and d must then be equal to nine. The only ways
to make the last digit of the product of two integers 9 would
be if those integers were either of:

1 and 9
3 and 3
7 and 7

So, both numbers must start with 9, end with either 1, 3, 7,
or 9, and one of them must be divisible by 11. The only
numbers divisible by 11 from 900 - 1000 are:

902
913
924
935
946
957
968
979
990

You can discard all of those that do not end in either 1, 3,
7, or 9, and you are left with:

913
957
979

So now the presumed answer is either:

(900 + 10 + 3)(900 + 10x + 3)
(900 + 50 + 7)(900 + 10x + 7)
(900 + 70 + 9)(900 + 10x + 1)

Factoring all those out, you get:

810000 + 9000x + 2700 + 9000 + 100x + 30 + 2700 + 30x + 9
824439 + 9130x

Now, for the first digit 824439 + 9130x to be 9, x must be 9
(if x were 8, then 824439 + 9130x = 897479, and the first
digit is 8). And so you have 913 * 993, which is the answer.
You can factor the others out to see if they produce a bigger
answer, which they don't.

Puzzle 3 without cheating

I am about 2800 times slower than the cheat I downloaded. But I get the right answer, and in a tolerable time (30 seconds). Pure brute forcing this seems like it would take much longer.

from sys import argv
from math import sqrt
from time import time

# PRIME GENERATOR

def nextprime():
    """ generate next prime """

    myprimes = [2]
    candidate = 3
    yield 2
    while True:
        factors = factorize(candidate,myprimes)
        if len(factors.keys()) == 1:
            myprimes.append(candidate)
            yield candidate
        candidate += 2

def primesupto(upto):
    """ collect primes from the generator"""

    result = []
    primes = nextprime()
    p = None
    while not p or p < upto:
        if p:
            result.append(p)
        p = primes.next()
    return result

def factorize(target,allprimes=None,verbose=False):
    """ get factors of an integer

    this works but is very slow if the primes aren't cached """

    factors = {}
    upto = int(sqrt(target))
    residual = target
    prime = nextprime()

    if allprimes:
        prime = allprimes.__iter__()

    while True:  # get factors and multiplicities up to sqrt(N)
        try:
            factor = prime.next()
        except StopIteration:
            break
        if factor > upto:
            break
        while not residual % factor:
            residual /= factor
            factors[factor] = factors.get(factor,0) + 1           

    if residual > 1: # if anything is left it is the sole prime factor > sqrt(N)
        factors[residual] = 1

    return factors
   
if __name__ == "__main__":
    target = int(argv[1]) # get target from command line
    upto = sqrt(target)
    starttime = time()
    allprimes = primesupto(upto)
    print max(factorize(int(argv[1]),allprimes).keys())
    endtime = time()
    print endtime - starttime

Update:

Best answer, wow!

from time import time
from sys import argv

n = int(argv[1])
d = 3

starttime = time()
while (d * d < n):
   if n % d == 0: n /= d
   else: d += 2
endtime = time()
print n
print endtime - startime

Runs in 0.7 msec. I feel like a complete idiot now. Oh well, it was interesting anyway.

Cheating, sorta

What is the largest prime factor of the number 600851475143 ?

from os import system
from os.path import isfile
from sys import argv

try:
  import numbthy
except:
  if not isfile("numbthy.py"): # don't mess with it if it's already there and broken
     system("wget http://userpages.umbc.edu/~rcampbel/Computers/Python/lib/numb
thy.py")
  import numbthy

f = numbthy.factors(int(argv[1]))
if f: print f[-1]

Level 2

I'm mostly at level 2 in this chart, except for IDEs. I don't really use them. I like aquamacs for an editor but even there I'm not good at it. I'm also pretty lame at version control, never having had the opportunity to work closely with others; hence never having run into a merge problem in real life.

I read some of the sorts of books he calls level 3.

Number Two

Python shines on this one. Find the sum of all the even-valued terms in the Fibonacci sequence which do not exceed four million.

from itertools import ifilterfalse

def fib(n1=1,n2=1,fmax=4000000):
    "terminating fibonacci sequence"

    assert 0 < n1 <= n2 < fmax
    l = [n2,n1]
    while l:
        if l[0] < fmax/2:
            l.append(sum(l))
        yield l.pop(0)

if __name__ == "__main__":
    print sum(ifilterfalse(lambda x: x % 2,fib()))

Euler Project

prob 1

A one-liner.

>>> sum([i for i in range(10) if not (i%3 and i%5)])
23
>>> sum([i for i in range(1000) if not (i%3 and i%5)])
233168

So it's surprisingly interesting:

>>> for n in range(1,9):
...     print sum([i for i in range(10**n) if not (i%3 and i%5)])
... 
23
2318
233168
23331668
2333316668
233333166668
23333331666668
2333333316666668

Can't even come up with a succinct notation for the theorem, never mind any idea how to prove it.

Moving notes to self

Notes to self or those interested in my daily pain will be moved to http://aardvarknoplay.blogspot.com/ . Old notes being moved.

This blog will actually have content that others may find interesting.

Some Key Papers in Climate Modeling History

A first pass at a greatest hits list; updates will not necessarily be noted as such. I had a request for key papers in the history of modeling so the emphasis leans that way. Nominations welcome.

Charney, Fjortofft & von Neumann 1950: Numerical integration of the barotropic vorticity equation, Tellus 2:237-254

Lorenz 1963: Deterministic Nonperiodic Flow J. Atmos. Sci 20:130-141

Lorenz 1967: The nature and theory of the general circulation of the atmosphere. WMO. http://eapsweb.mit.edu/research/Lorenz/General_Circ_WMO_1967_Part1.pdf

Manabe & Bryan 1969: Climate calculations with a combined ocean-atmosphere model.

Bryan & Cox 1972: The circulation of the world ocean: a numerical study. Part I, a homogeneous model J Phys Oceanog
Arakawa & Lamb 1977: Computational Design of the basic processes of the UCLA general circulation model. Methods in Computational Physics 17:173-265

Charney et al 1979: Ad Hoc Study Group on Carbon Dioxide and Climate. http://www.atmos.ucla.edu/~brianpm/charneyreport.html

Manabe & Stouffer 1988: Two stable equilibria of a coupled ocean-atmosphere model

Ramanathan et al 1985: Trace gas trends and their potential role in climate change. J Geophys Res 90:5547-5566

Cess et al 1990: Intercomparison and interpretation of climate feedback processes in seventeen atmospheric general circulation models

Bryan 1991: Poleward heat transport in the ocean: a reviwe of a hierarchy of models of incerasing resolution. J Phys Oceanog 18:851-867

Cubasch et al 1992: Time-dependent greenhouse warming computations with a coupled ocean-atmosphere model

Hansen et al 1992: Potential climate impact of Mount Pinatubo eruption GEOPHYSICAL RESEARCH LETTERS, VOL. 19, NO. 2, PP. 215-218, 1992
doi:10.1029/91GL02788

Hasselmann 1997: Multi-pattern fingerprint method for detection and attribution of climate change. Climate Dynamics 13:601-612

Hansen et al 1998: Climate Forings in the Industrial Era. PNAS

Bengtsson 1999: From short-range barotropic modeling to extended-range global weather prediction. Tellus 51 A-B:13-23

Randall 2000: General Circulation Model Development: Past, Present, and Future (International Geophysics) [Hardcover]

Hansen et al 2005: Earth's Energy Imbalance: Confirmation and Implications. Science.

Hack et al 2006: Simulation of the Global Hydrological Cycle in the CCSM Community Atmosphere Model Version 3. J Clim Volume 19, Issue 11 (June 2006)

Schmidt et al 2006: Present-Day Atmospheric Simulations Using GISS ModelE: Comparison to In Situ, Satellite, and Reanalysis Data. J. Climate, 19, 153–192.

Rahmstorf et al 2007: Recent Climate Observations Compared to Predictions. Science 316:709 DOI: 10.1126/science.1136843

More Adventures in Climate Analysis Installs

So here I document getting Ferret to work on a Mac; getting it to run on the super is another problem. I assumed no privileges and more or less succeeded. However, the install instructions are (as is normal with scientist code) broken.

Here are the absurd installation instructions:

http://ferret.pmel.noaa.gov/static/Downloads/mac_intel_downloads.html

This is basically just three useful links to the actual binaries under "Environment and Executable files"; ignore the rest of it.

Then proceed to

http://ferret.pmel.noaa.gov/static/Downloads/ferret_installation_and_update_guide_v600.html

If you don;t knwo about environment variables in bash, go study up on them.

The first step 1 should be

tar xf $GET_LOCATION/fer_environment.tar

The second step 1 and step 2 [sic] work as noted. Step 3 should be

tar xf $GET_LOCATION/fer_dsets.tar

as the Mac ungzips behind your back. Your next step will have to be backwards, as the Finstall script expects the exectuables to be gzipped. So go back to your download directory and type

gzip fer_executables.tar

Also, you will have to find a place for an extra file in the middle of the next step. Be sensible, create a ~/.Ferret directory before doing the next step. Use that for your $SET_FER .

Now you can run the Finstall per instructions. But the script you ghet out will be a csh script not a bash script. The following edits worked for me:

export FER_DIR=/Users/tobis/Ferret
export FER_DSETS=/Users/tobis/FerSamp

#   if ( $PATH !~ *ferret* ) then
#       export PATH={$PATH}:$FER_DIR/bin
#   else
#        echo replacing ferret path
#      export PATH=`echo $PATH | awk -F: '{for (i=1; i<=NF; i++) {if ($i \!~ /ferret\/bin/) {printf "%s:",$i}}}'`
#      export PATH={$PATH}$FER_DIR/bin
#   endif
 
# =========== Initially make no modifications below this line ===========

#  Default Ferret document browser
#   export FER_WEB_BROWSER="netscape -ncols 60"
   export FER_EXTERNAL_FUNCTIONS="$FER_DIR/ext_func/libs"
   export FER_GO=". $FER_DIR/go $FER_DIR/examples $FER_DIR/contrib"
   export FER_DATA=". $FER_DSETS/data $FER_DIR/go $FER_DIR/examples $FER_DIR/contrib"
   export FER_DESCR=". $FER_DSETS/descr"
   export FER_GRIDS=". $FER_DSETS/grids"
   export TMAP="$FER_DIR/fmt"
   export PLOTFONTS="$FER_DIR/ppl/fonts"
   export SPECTRA="$FER_DIR/ppl"         # for old ferret versions
   export FER_PALETTE=". $FER_DIR/ppl"       # palette search list

So basically, change all the setenvs to exports, put in the equal sign, comment out the if/else block. If you are better with bash scripts, the if/else thing looks harmless.

I did not uncomment the reference to netscape. I don;t know where to get a netscape for os X 10.6 anyway. Sigh.

This actually works!

Meanwhile Chas has bunches of hoops for me to jump through. Will keep you posted, if you exist.

PyNGL Even worse than CDAT?

Undecided as yet. More to follow. But plenty of the same signs of cluelessness.

I am still hopeful that the interpolation routine I need will work. And the install is much more straightforward, there being suitable binaries for the machines I care about.

But the test routine starts with an absolute path to a python installation at NCAR!

And this is unpythonic, though perhaps this is a Fortran 77 constraint rearing its ugly head

import Ngl
import Nio
import numpy

# create an array and make a netcdf file

ff = Nio.open_file("foo.nc","c")
ff.create_dimension("x",10)
ff.create_dimension("y",10)

ff.create_variable("u","f",("x","y"))

z = numpy.zeros((10,10),"f")

# ff.variables["u"] = z doesn't work

ff.variables["u"].assign_value(z)

# assignment is by value, as the name indicates

z[3,3] = 3.3

ff.variables["u"][4,4] = 4.4

# but after assignment you can pick up a reference

u = ff.variables["u"]

u[2,2] = 2.2

"""
>>> ff.variables["u"][:]
array([[ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  2.2,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  4.4,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
      [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.]], dtype=float32)
"""

ff.close()

Anthony Doerr

Read the whole thing. It starts

During my sophomore year, 1992, 1,500 scientists, including more than half the living Nobel laureates, admonished in their Warning to Humanity: “A great change in our stewardship of the earth and the life on it is required if vast human misery is to be avoided and our global home on this planet is not to be irretrievably mutilated.”

So what have we done? Not much. From 1992 to 2007, global CO2 emissions from burning fossil fuels rose 38 percent. Emissions in 2008 rose a full 2 percent despite a global economic slump. Honeybees are dying by the billions1, amphibians by the millions, and shallow Caribbean reefs are mostly dead already.2 Our soil is disappearing faster than ever before, half of all mammals are in decline, and a recent climate change model predicts that the Arctic could have ice-free summers by 2013. Unchecked, carbon emissions from China alone will probably match the current global level by 2030.

“The god thou servest,” Marlowe wrote in Dr. Faustus, almost four hundred years before the invention of internet shopping, “is thine own appetite.” Was he wrong? How significantly have you reduced your own emissions since you first heard the phrase “climate change?” By a tenth? A quarter? A half? That’s better than I’m doing. The shirt I’m wearing was shipped here from Thailand. The Twinkie I just ate had 37 ingredients in it. I biked to work through 91-degree heat this morning but back at my house the air conditioner is grinding away, keeping all three bedrooms a pleasant 74 degrees.

My computer is on; my desk lamp is glowing. The vent on the wall is blowing a steady, soothing stream of cool air onto my shoes.

h/t Andrew Sullivan. Anthony Doerr, whom I had not heard of until today, lives in Idaho, writes "on Science" column for the Boston Globe, and is a 2010 Fellow of the John Simon Guggenheim Memorial Foundation.

Exploring NIO

The PyNIO install from binary went more or less without incident; I had to dig up a conformant numpy (1.5 is out but the binary required 1.4)

I am deferring a PyNGL install until I figure out what PyNIO can do.

So now I have py 2.5.1, numpy 1.4.1, Nio 1.4.0 and a doc.

The main doc is here

Not hard for me to dig up a netcdf file from the CAM distribution; let's take a 2 D dataset from the land surface model. No idea what is in it.

-rwxr-x---  1 tobis G-25522 7671124 Oct 22  2009 clms_64x128_USGS_c030605.nc

seems to work! The data object has some attributes set by the data file.

>>> import numpy as np
>>> import Nio
>>> data = Nio.open_file("nctest.nc")
>>> data

>>> dir(data)
['Conventions', 'Glacier_raw_data_file_name', 'History_Log', 'Host', 'Inland_water_raw_data_file_name', 'Input_navy_oro_dataset', 'Lai_raw_data_file_name', 'Logname', 'Revision_Id', 'Run_mode', 'Soil_color_raw_data_file_name', 'Soil_texture_raw_data_file_name', 'Source', 'Urban_raw_data_file_name', 'Vegetation_type_raw_data_filename', 'Version', '__class__', '__delattr__', '__dict__', '__doc__', '__getattribute__', '__hash__', '__init__', '__module__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__str__', '__weakref__', 'close', 'create_dimension', 'create_variable', 'set_option']

but how to get at the actual contents?

Ah; the directory is incomplete. Apparently the extension module wasn't Pythonic. This can be patched.

So we can get at dv = data.variables() yielding

['PCT_CLAY', 'LATIXY', 'LANDMASK', 'LANDFRAC_PFT', 'PCT_LAKE', 'LANDFRAC', 'NUMLON', 'LONGXY', 'MONTHLY_HEIGHT_BOT', 'PCT_WETLAND', 'MONTHLY_SAI', 'PCT_URBAN', 'PCT_SAND', 'MONTHLY_HEIGHT_TOP', 'PCT_PFT', 'MONTHLY_LAI', 'SOIL_COLOR', 'PCT_GLACIER', 'PFT']

Now, just printing (__str__) data tells us

   dimensions:
      lsmlon = 128
      lsmlat = 64
...
   variables:
...
      float PCT_LAKE [ lsmlat, lsmlon ]

so dv['PCT_LAKE'] ought to be a 128 x 64 array

but dv[2000,0] works. Apparently an overflow is treated as a [-1] index and an underflow as a [0] index. Not very reassuring.

Exasperatingly, though PyNIO is available for py2.5 and numpy 1.4, PyNGL is not. Must get python 2.6... boring...

Install confusion

Trying to install PyNCL on a Red Hat machine without sudo priv's.

1) Py2.7 from source. Builds fine. Runs fine.

2) numpy from source using aforesaid python. atlas/blas instructions aren't too clear, proceed blindly ahead. build works fine.

3) import numpy or import numarray fails:

ImportError: Error importing numpy: you should not try to import numpy from
its source directory; please exit the numpy source tree, and relaunch
your python intepreter from there.

very little info on Google about this. Try various versions of current working directory. Failure all the same.

4) looks like this code:

if __NUMPY_SETUP__:
    import sys as _sys
    print >> _sys.stderr, 'Running from numpy source directory.'
    del _sys
else:
    try:
        from numpy.__config__ import show as show_config
    except ImportError, e:
        msg = """Error importing numpy: you should not try to import numpy from
        its source directory; please exit the numpy source tree, and relaunch
        your python intepreter from there."""
        raise ImportError(msg)
    from version import version as __version__

is getting invoked. Try setup command, but it tries to install in /usr/local/lib/python2.7

5) How to tell setup where to put the numpy?

setup.py install --help

helps, yielding

Options for 'install' command:
  --prefix            installation prefix
  --exec-prefix       (Unix only) prefix for platform-specific files
  --home              (Unix only) home directory to install under
  --install-base      base installation directory (instead of --prefix or --
                      home)
  --install-platbase  base installation directory for platform-specific files
                      (instead of --exec-prefix or --home)
  --root              install everything relative to this alternate root
                      directory
  --install-purelib   installation directory for pure Python module
                      distributions
  --install-platlib   installation directory for non-pure module distributions
  --install-lib       installation directory for all module distributions
                      (overrides --install-purelib and --install-platlib)
  --install-headers   installation directory for C/C++ headers
  --install-scripts   installation directory for Python scripts
  --install-data      installation directory for data files

It must be one of those!

6) Used "--prefix". setup.py runs without complaint.

Haha! New error. "{PATH}/multiarray.so: cannot open shared object file: No such file"

but the file exists! Now what?

7) Wait - no. I used the system python not the self-installed 2.7. It works! numpy is imported.

8) Now get PyNCL? (ncl is already working on this machine)

NO! Looks like I have to do the numpy thing again, because PyNGL recommends against the g77 compiler and requires gfortran. So my numpy, though working, is wrong.

9) While I'm at it, no BLAS/ATLAS? etc?

The Russia Thing

I am not sure how to take being definitely regarded as climate scientist by both Roger Pielke Jr (who is in and out of the Tobis-ooh-scary club) and Brad Johnson (a Romm protege), but not quite so definitively by anyone commonly regarded as a climate scientist. I suppose I should accept it as better than nothing. But Johnson's topic is worth following up. And I really liked that Brad got an actual meteorologist to back me up:

Like Dr. Tobis, Carver [Meteorologist Rob Carver, the Research and Development Scientist for Weather Underground] believes that manmade global warming has fundamentally altered weather patterns to produce the killer Russian heat wave. “Without contributions from anthropogenic climate change,” Carver said in an email interview with the Wonk Room, “I don’t think this event would have reached such extremes or even happened at all”:

I agree with Michael Tobis’s take at Only In It For the Gold that something systematic has changed to alter the global circulation and you’ll need a coupled atmosphere/ocean global model to understand what’s going on. My hunch is that a warming Arctic combined with sea-surface-temperature teleconnections altered the global circulation such that a blocking ridge formed over western Russia leading to the unprecedented drought/heat wave conditions. Without contributions from anthropogenic climate change, I don’t think this event would have reached such extremes or even happened at all. (You may quote me on that.)

Cool. Thanks for taking the heat, making a stronger claim than I did, Rob! And thanks for the work on this, Brad!

I want to explain my thinking in some detail on this, but I really don't have the time to do it justice right now. Of course it will be half-wittedly nitpicked and mocked to oblivion if I post it on a blog whether it's brilliant or foolish. I wonder if it rates a real paper somewhere. That would be a novelty...

But a short version occurs to me. Remember when I tried to explain that "global warming" and "anthropogenic climate change" don't mean exactly the same thing? Well, the fact that we now have situations like this one, and last year's in Australia, allows for making the point clear with some examples.

See, what we are worried about is not global warming. Global warming itself causes relatively little damage, at least at first. What we are worried about is climate change.

Anthropogenic global warming does not cause climate change.

Anthropogenic climate change causes global warming,.

Global warming (at the surface, just like global cooling in the stratosphere) is one of the more predictable symptoms of anthropogenic climate change. But if you change the forcing, you change the response. That's not a very deep scientific result, is it?

"Anthropogenic climate change" - well that's a tedious mouthful, isn't it? Not in a sense, no, not really. Not as tedious as what is happening this instant in Pakistan and in Russia, anyway. That's what abrupt climate change looks like.

Handy FIR filter inherits from list:

   class filter(list):

   """real valued F I R filter with memory"""

   def __init__(self, coeffs, data=None):
            list.__init__([float(item) for item in coeffs])
       if data:
           assert len(data) == len(coeffs)
           self.data =[float(datum) for datum in data]
       else:
           self.data = len(self) * [0.]

   def __call__(self,value=0.):
       return self.filter(value)

   def output(self):
       return sum(map(mul, self.coeffs, self.data))

   def update(self,value):
       self.pop(0)
       self.append(value)

   def filter(self, value=0.):
       value = float(value)
       self.update(value)
       return self.output()

   def innovate(self):
       return self.filter(white())