Ox 3.00: new features and fixes

New features and enhancements

New functions

insertc, insertr, setbounds, outer, diagcat
rangh, rangig, ranshuffle, ranpoissonprocess, ranbrownianmotion
probbvn, probmvn (bi- and tri-variate normal probabilities)
tailf, tailt (non-central F, non-central t distributions), densgh, densgig
SolveQP (quadratic programming)
maxc, maxcindex, minc, mincindex, sortcindex
new graphics functions: DrawAcf, DrawHistogram, DrawXYZ; extended: DrawDensity
dayofweek, dayofcalender
chdir, getenv, fremove, getcwd, systemcall
bessel (fractional orders), dawson, expint
strfind to find character in a string, strfindr for reverse searching, also to search for substring in a string; strifind and strifindr are case insensitive versions.

Enhanced functions

probt, tailt, probf more accurate for small probabilities. probbeta, tailt, probf more accurate for large a,b. Non-central distributions more accurate over larger range.
probt, tailt, quant with non-integer df.
decqrmul now works for reversed H; one argument version to compute Q. decqr now allows m < n. decldlband now sets diagonal of L to 1.
idiv and imod can handle matrix arguments
vecfind/vecrfind to search for specific value now work with missing values.
Can write dht/dat file; use fread/fwrite on dht files; ability to fread 4-byte reals.

Syntax enhancements

#import can now also import .ox files. It will first try to find the .oxo file, and if that fails it will search for the .ox file.
Ox can now run many Gauss files directly. Use oxl -g from the command line or OxGauss with Ox Professional.
Now allowing decl in function header; also allowing declaration in for loop, as in:
for (decl i=0;...) s
which equals
{decl i=0; for (;...) s }
indexing string for reference: can index one beyond end, which returns 0.
Optional warning for accessing matrix as vector (see oxwarning function)
Parser recognizes .NaN and .Inf; relational operators now treat missing values properly.
new array[r][c] and new matrix[r] now allowed.
Moved power to above unary ops, so -x^2 is now identical to -(x^2). This change only affects code where the power is even: previously, the sign would disappear, but now the sign of the result is changed.
x[][] and x[1][] now works for 1x0 matrix.
concatenation for illegal types (functions etc.): casting lhs to array, so that concatenation will work.

Run-time system

Moved to OX3PATH environment variable instead of OXPATH. Setting default OX3PATH value under Windows based on location of oxwin.dll. Therefor installation program does not need to set OX3PATH, nor update the path anymore.
Increased default stack size to -s2000,1000

Ox classes

Simulation class: added support for two-sided tests and saving of results.
Numerous enhancements to Modelbase and OxPack.
Several small labelling changes to Simulation/PcNaive/PcFimlDgp class for PcNaive.
Added UnitRoot test to PcFiml class with Ericsson&MacKinnon 5%/1% critical values.

New Packages

Financial Numerical Recipes: Financial Numerical Recipes is an Ox port of the code written by Bernt Arne Ødegaard for advanced financial calculations. It contains the basic and some advanced algorithms for option pricing, and some algorithms dealing with term structure modeling and pricing of fixed income securities.
Lapack: Lapack is a package for solving linear equations, linear least squares problems, eigenvalue problems, and singular value problems. This is a port to Ox of the main driver files for real and symmetric matrices from lapack, part of netlib.
Loess Package: Loess is a package for smoothing of multivariate scattered data (LOESS); decompose time series into trend + seasonal + remainder (STL). This is a port to Ox of the port to Ox of the loess and stl code in netlib.

Problems in version 2.20 which have been fixed in version 3.00:

cumulate(rann(0,1)) would crash
CloseDrawWindow via OxRun did not clear graphics window
discretize on empty matrix crashes
problem with decqr if m < n
rank of empty matrix gave runtime error
SetDraw(SET_PRINTPAGE, PAGE_USER, PAGE_LANDSCAPE, 600, 1300); had documentation error, and would set both X and Y size to 1300.
sumsqrc/sumsqrr of integer/double scalar would not take square. Integer power of negative integer could overflow.
fixed minor problem with loading Excel 97/2000 files.
fixed rare instance of infinite loop in axis division
LoadDta didn't set specified sample
string == "" could crash when running oxo file
Sample::ObsYear and Sample::ObsPeriod incorrect for some negative indices.
print("%% crit") now prints one %
x[][:-1] returns whole matrix, but should give index error.
int==string returns a value, but should fail like string==int
Using '\\' in code which is skipped because of preprocessor statement would lead to scan errors.
Fixed sign problem with eigenvectors of general matrix when there are isolated eigenvalues.
betafunc(1,a,b) was wrong
matrix[1][1] * matrix[r][c] didn't work
if (x[0]) wouldn't dereference one index, and always be true.
In sscan("9", "%2d", &i), the variable i would be set to 9. However, in sscan("09", "%2d", &i) "09" is treated as octal because it is prefixed with a zero. Now "i" is treated as "d", and x,u,o formats are also allowed.
as=as[2:3] would not work for arrays (but as1=as[2:3] world work). Also solved problem with indexing of higher order arrays.
This doesn't parse: println(!!2);
Fixed densgamma for x --> 0 for nu <= 1