Changes in 3.1.0.0
Changes in 2.6.0.0
1. Further changes to parser/interpreter:
(a) rlab-3.1 and rlab-2.6: classdef(...){...} environment accepts static(...) declaration for class-member private
properties and methods.
(b) rlab-2.6 differs from rlab-3.1 ONLY in if-else statement:
rlab-2.6: if (cond) { ... else ... } [full support for legacy scripts]
rlab-3.1: if (cond) { ... } else { ... }
2. Continuing work on cleaning/leaning/deglitching the code.
3. Miscelaneous improvements to existing functions.
[PREVIOUS POSTS]
Changes in 3.1.0.0
1. Further changes to parser/interpreter:
(a) classdef(...){...} environment with private(...), public(...) and instantiation variables
that allows programmatic creation of lists (class members). The list entries
can be either variables (public properties) or functions (public methods), where
all so created lists (class members) share constructor-static common entries
which can be either variables (private properties) or functions (private methods).
Their syntax and usage partially follows C++ (in that :: is not supported).
(b) Special system variable 'ans' is distinguished from operator '?':
- ending statement without ';' causes it to be assigned to 'ans', and then printed. This
may be convinient for command-line computations;
- ending statement with '?' causes it to be only printed.
(c) Strings can now be defined over multiple lines: parser looks for pair of
matching " "
(d) Internal fixes that prevent segfaults in case of scripts full of errors.
Changes in 3.0.6.0
1. Major changes in PARSER which make rlab3 not-compatible with rlab2:
(a) New format of the if-else-command follows c-language:
if (cond) { ... }
if (cond) { ... } else { ... }
if (cond) { ... } else if (cond) { ... } ...
rather then the old-style
if (cond) { ... else ... } [ rlab2 ]
NB: To change your files to new format, replace 'else' with '} else {', e.g., using
the following script ONCE:
# copy old file with extension .r to files with extension .r3
rename .r .r3 *.r
# change else command in .r3 files
sed -i -e 's/ else/ } else {/g' *.r3
(b) added for-then and while-then constructs that follow the python's for-else
and while-else.
(c) added do-while/until loop that follows c-language (do-while)
do { ... } while/until (cond);
(d) added switch command that follows c-language definition, with that difference
that part of the case statement can be an expression or variable.
(e) added element-wise operations +=, -=, *= and /= mostly without storage of
intermediate results.
(f) static declaration of variables allowed inside function body, which specifies
function-static variables, which values are kept between function calls.
(g) EOF token that stops processing of the current file and returns control either
to calling file or to the prompt.
(h) script{} environment that allows programmatic manipulations of scripts.
2. rlab3 is configured so it can coexist with rlab2 on the same system.
The rlab2 keeps the extension "*.r" , while rlab3 takes on ".r3"
Changes in 2.5.0.3
1. Improved configure script. Unneeded libraries are removed from configuration.
Changes in 2.5.0.2
1. Continuing work on cleaning/leaning/deglitching the code. Improved build script so it reacts
properly to two different versions of plplot library (libplplot.so vs. libplplotd.so).
2. Fixes of glitches in plplot interface. Better documentation in rlabplus manual.
3. Added Powell optimization solvers newuoa (unconstrained), bobyqa (simply constrained),
lincoa (linear constraint), and cobyla (functional constraint).
Changes in 2.5.0.1
1. Continuing work on cleaning/leaning/deglitching the code.
2. Improved interface to plplot - now it is more gnuplot-like within rlab environment.
3. Miscelaneous improvements to existing functions, e.g., hist().
Changes in 2.5.0.0
1. EXPERIMENTAL New data type MATRIX_DENSE_ENTITY, or 'cell' (matrix-like collection of rlab objects, accessed by
the indices, just like regular matrix) that is an alternative to
'list' (unstructured collection, accessed with a key);
2. Continuing work on cleaning/leaning/deglitching the code: e.g., readm.
Changes in 2.4.2.5
1. Continuing work on cleaning/leaning/deglitching the code.
Changes in 2.4.2.4
1. Continuing work on cleaning/leaning/deglitching the code: ImageMagick interface.
2. Improved command line interface to rlab and scripts. More options for serial port.
Changes in 2.4.2.3
1. Continuing work on cleaning/leaning/deglitching the code. Overhaul of some functions with more options
added, c.f., floor, ceil, mean, var, covar, polyval, polyder.
2. Modifications of parser: now accepts comments during a multi-line definition of an array. The scripts
relying this feature may be incompatible with older versions.
Changes in 2.4.2.2
1. Continuing work on cleaning/leaning/deglitching the code.
2. Further work on parser: Now it recognizes c-style comments /* ... */, that is, ignores them.
3. Minor fixes in many libraries (libooo, libsystem, libstdio, to mention a few) so that they
better handle error conditions.
4. String functions now fully integrate Pattern matching, cf., findstr, strindex(r), strsplt, strtod,
lstrip/rstrip/strip, gsub, grep, readm, reads.
Changes in 2.4.2.1
1. Experimental implementation of string functions based on Pattern matching library.
Changes in 2.4.2.0
1. Continuing work on cleaning/leaning/deglitching the code.
2. Added feature to parser that recognizes integers: in hexadecimal format using prefices "0x" or "0X,"
e.g., 0x034; in binary format using postfices "B" or "b," e.g., 10010101B; or as plain integer using postfices
"l" or "L," e.g., 47891491741L. This may make scripts for latest versions incompatible with the older versions.
E.g., a statement 'x=0xaa' creates an integer variable which content is 170 in decimal notation. Previously
this could have been accomplished through sequence of statements 'x=int(strtod("0xaa"))'
3. added some manipulations with integers, e.g., bit-rotation and bit-shift.
4. expanded h5 family of functions.
Changes in 2.4.1.2
1. Continuing work on cleaning/leaning/deglitching the code.
2. Reworked neural network solvers that used to be tied to (now extinct) library sprann. These are now
hooked up to "c-library" for pattern recognition.
Changes in 2.4.1.1
1. Fixed segfault occuring on some systems caused when parser would encounter error in the script. Partial fix
segfaults as a result of repeated errors that occur when the 'eval' function has improper argument, e.g., evaluate value
of expression in which there are undefined variables, >> clear(x); eval("x+6")
2. Further fixing of the scanner: changed type of 'nargs' within function call from 'double' to 'MDR' so that
it can be operated on.
3. Modified 'make' script so that internally 'make scanner' preceeds 'make rlab'.
4. Continuing work on cleaning/leaning/deglitching the code.
Changes in 2.4.1.0
1. Fixed segfault occuring when calling solvers with (functions with no arguments) as arguments.
2. Fixed backward compatibility with gsl (e.g., robust linear fit was introduced recently, rlab can be linked to
old versions of gsl).
Changes in 2.4.0.0
1. Thorough internal reorganization of solvers that call functions as arguments. A number of mini memory leaks
have been resolved in the process. Certain invocation procedures modified so their execution is 10-20% faster.
Haxes replaced by procedures uniform across different solvers.
Changes in 2.3.3.8
1. Continuing work on cleaning/leaning/deglitching the code. Case in point: preventing existing functions
to cause segmentation fault or return non-sensical results if their arguments were not correct.
E.g., y = ifelse(x==[], 1, x[1]) used to fail if x is empty matrix because the function would assume that
all three supplied arguments exist (here, x[1] does not exist if x is an empty matrix).
E.g., polyval(x,a) treats properly any vector 'a', rather then just row vector. Previously, if 'a' was column
vector it would go internally out of bounds, and return non-sensical result.
Please note, in helping me eliminate such glitches I encourage you to send me nonsenses, or segfaults rlab
creates while running your scripts.
3. More work on incorporating NaN's in rlab workings, e.g., see functions diff() or quot().
4. Fixed functions writem() when operating on hdf5 file in append mode: it deletes previous value
of the object before entering new one. Also writem("somefilename.h5", [], "/hdf/abs/object/name") will delete
object "/hdf/abs/object/name" if argument is empty matrix.
Changes in 2.3.3.7
1. Continuing work on cleaning/leaning/deglitching the code.
Changes in 2.3.3.6
1. Continuing work on cleaning/leaning/deglitching the code.
2. More work on patterns in function 'compact', set operations now work with string vectors as well.
3. sorting function has now three ways of dealing with nans: put them on top of the list, on bottom,
or leave them in place.
Changes in 2.3.3.5
1. Continuing work on cleaning/leaning/deglitching the code.
2. Some work on patterns is implemented in the function 'match'
3. Added operation with sets, some functionality with tables. All can tolerate nans.
Changes in 2.3.3.4
1. Fixed corrupted 'y.tab.h' file in 2.3.3.3.
2. Continuing work on cleaning/leaning/deglitching the code.
Changes in 2.3.3.3
Changes in 2.3.3.2
1. Continuing work on cleaning/leaning/deglitching the code.
2. Modified build script so that flex and bison can be used through 'make scanner'.
Updated default rlab parser to output of flex 2.5.37.
Changes in 2.3.3.1
1. Continuing work on cleaning/leaning/deglitching the code.
Changes in 2.3.3.0
1. Consolidated rlab complex codes.
2. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code: case in point harminv and time-frequency toolbox.
Changes in 2.3.2.3
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
2. Fixed build script: It can recognize libngspice on 32-bit systems. It builds the .bin. version
without any addons with pgplot for plotting - see ./rconfigure_low32 in source root directory.
Changes in 2.3.2.2
1. Added suggested build script for Raspberry Pi, "rconfigure_rpi." Rlab tested and operational
on Raspberry Pi running opensuse 13.1. Binary version available, as well.
Fixed different "#define" statements so that the same sources can be built on either
of three architectures, armv6l, x86 & x86_64.
2. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
Changes in 2.3.2.1
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code: emphasis on date time funcions.
Changes in 2.3.2.0
1. New internal type: dense matrix, existing dense matrix formats converted to dense matrix type.
2. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code. Statistical functions improved so to ignore nan's.
3. Improved plplot interface: it now accepts lists for plotting 3-d lines and surfaces (plsplot)
"mesh" style or "3d" style, and contours (plcont). Improved interna efficiency of data transfer
to plplot.
4. Pattern recognition functions added: classify, cluster, distance matrix, localabsmaxi, localrelmaxi.
Changes in 2.3.1.8
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
2. Updated configure script so to handle properly ImageMagick and gphoto2 library configurations.
User still has to request the feature during configuring.
Changes in 2.3.1.7
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code, e.g., prompt, or 'ans'.
Changes in 2.3.1.6
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code. Silent upgrades completed.
Changes in 2.3.1.5
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
Changes in 2.3.1.4
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
2. Deglitched extraction of rlabplus-addons, so that scaled down version of rlab+rlabplus compiles
even on old 32-bit machines (currently oldest machine that runs rlab is a
toshiba laptop 1GHz, 512MB RAM, with suse 11.1 on it). Its configuration is given in
script 'myconfigure_low32'.
3. Please note: All binaries available on the sourceforge are built using 'myconfigure_low32'.
For all rlab+rlabplus features one has to build from sources.
Changes in 2.3.1.3
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code: fixed some internal lack of features and overly optimistic dealing with non-existent
variables.
2. Fixed omission in 'configure' script so that location of the jvm is properly passed to
rlab start-up script. Same with some other libraries.
3. Finished filter. It works now for complex numbers as well.
Changes in 2.3.1.2
1. Changes in configuration of rlab.
I was hit by blast from the past: most of special-topic libraries added to rlab as a part of
rlabplus are now optional, and need to be specified during 'configure' to be included.
This was done so that rlab can be built on very old laptops running even older linuces
(do you remember kde3.5?). What is still required are: hdf, gsl, curl shared library.
On older systems arpack, glpk and superlu might be available as static libraries. Configure script
takes that into account: One still needs to specify only the directory where the library resides,
and hope that the headers are available through /usr/include.
Overall, these are the latest flags needed to build rlab totally (see also 'myconfigure' script that
lists them all):
./configure
--with-jvm=[where is my JVM located? 'myconfigure' can find it for you]
--with-ngspice=[where did I put shared ngspice when I built it? 'myconfigure' can try to find it]
--with-arpack[=/dir/to/static or shared library, but only if different from default locations]
--with-superlu[=/dir/to/static or shared library, but only if different from default locations]
--with-im[=/dir/to/shared library, but only if different from default locations]
--with-gphoto2[=/dir/to/shared library, but only if different from default locations]
--with-plplot[=/dir/to/shared library, but only if different from default locations],
or --with-pgplot, for the latter rlab contains its source tree - needed for old systems without plplot
--enable-python[=/dir/to/shared library, but only if different from default locations]
--with-suitesparse[=/dir/to/shared library, but only if different from default locations]
--enable-glpk[=/dir/to/static or shared library, but only if different from default locations]
Default behavior is not to install any. Even then, one gets system with bunch of improvements and new
features, with many a bug or glitch fix.
Changes in 2.3.1.1
1. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code: fixed some HDF bugs and glitches: 'isfile' works, reading and writing string matrices works.
2. Additional features to existing solvers, e.g. mean/var. Fixed filter function: it can now take
the initial output as input parameter. Fixed 'eval' function: does not segfault if improper input.
3. Statistical lists <<val;wgt>> now support algabraic operations +, -, ./, .\, .*,
and *. Some functions, e.g., mean,var, ..., support them from before.
4. Fixed glpk so it properly compiles on older version of the library (parameters have changed recently).
5. Fixed rgphoto in regards to nonexistent uint8_t type.
6. Rebuilt the plplot interface, as any further work on pgplot has been ceased (it blows on xinerama setup,
creates windows that take half screen - 1 monitor). Most pgplot commands and structures should work under
plplot: the two have almost identical functions.
User now has to choose between --with-plplot and --with-pgplot flags when 'configure'-ing.
7. 'ans' variable stores the most recent result of command line calculations.
Changes in 2.3.1.0
1. Fixed build-up script and rectified some omissions.
2. Added libgphoto2 through new command list 'gp'. If not wanted
configure rlab with '--disable-gphoto2'
3. Added parts of ImageMagick through new command list 'image'. If not wanted configure
rlab with '--disable-im'
4. Updated GLPK wrapper to work with versions of GLPK 4.52 and higher: Do I need to put this
with --disable-glpk flag?
5. Cleaning/leaning/deglitching of the code. Minor fixes to some mathematic functions, e.g., mod.
6. If building from sources, there is a script 'myconfigure' which tries to figure out flags
for compilation in regards to python, java, spice.
Changes in 2.3.0.4
1. Fixed documentation omission. Flag for using java virtual machine should read
--with-jvm=/dir/where/yourlibjvm.so.is
and it should not contain the file name libjvm.so itself. Silly me. Also fixed some other
glitches in configure script.
2. Cleaning/leaning/deglitching of the code. Modified output of diffevol solver and added convergence
controls.
Changes in 2.3.0.3
1. Notable new solver: a version of differential evolution algorithm is now available as solver
'diffevol'.
2. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
Changes in 2.3.0.2
1. Build flags are better defined in configure script
--disable-python (by default it is enabled, provides command list 'py')
--with-jvm=/usr/lib64/libjvm.so (by default it is disabled, if enabled provides
command list 'jvm')
--with-ngspice (by default it is disabled, if enabled provides command list 'spice')
2. Added support for ngspice built as a shared library.
3. Updated to the latest version of the Boehm's Garbage Collector (gc-7.2).
4. Continuing work on adding more options to existing solvers, and cleaning/leaning/deglitching
the code.
Changes in 2.3.0.1
1. Removed static libraries arpack, amdpack, umfpack and superlu, and replaced them with
the shared object ones available from your linux distribution - check install.txt.
2. Continuing work on: adding more options to existing solvers and functions,
and cleaning/leaning/deglitching the code: fixed ngspice, added calendar library.
3. Add a calendar/date/time toolbox.
Changes in 2.2.12.16
1. Added jvm set of functions that allow rlab to call java methods and passes parameters to them and
get results from them.
2. Continuing work on: adding more options to existing solvers and functions,
and cleaning/leaning/deglitching the code.
Changes in 2.2.12.15
1. Added flag -e "cmds" so that commands can be executed when invoking rlab2 from the command line.
The flag parameter "cmds" does not allow white spaces (feature of GNU getopt).
2. Fixed 'configure' script so that gcc version is correctly recognized.
3. Continuing work on: adding more options to existing solvers and functions,
and cleaning/leaning/deglitching the code.
Changes in 2.2.12.14
1. fixed build script - glpk was improperly removed.
Changes in 2.2.12.13
1. libglpk removed from source. User needs to download the library and development files, and install them
on his/her own.
2. Fixed wrong location of rlab headers in python.c.in
3. Fixed glitch in main.c that appears in newer versions of gcc, related to 'yydebug'
Changes in 2.2.12.12
1. Fixed build script for python, so it builds properly on opensuse and
fedora systems (thanks a.g.).
2. Continuing work on: adding more options to existing solvers and functions,
and cleaning/leaning/debugging the code.
3. More functions, e.g., showkey, prompt, inputs.
Changes in 2.2.12.11
1. Embedded python interpreter is now available.
2. Added some more options to existing solvers.
3. Continued work on removing bugs, and buglets, and code cleaning.
Changes in 2.2.12.10
1. 'make flibs' now builds pgplot
2. fixed linking issues between 64 and 32-bit versions (each links to its respective
/usr/lib{64} libraries). Fixed variable in 'configure' so that the system architecture
is properly registered - should work on opensuse and redhat fedora systems.
3. continuous work on code cleaning and glitches removal.
Changes in 2.2.12.8
1. Fixed the configure/Makefile so it properly links to libraries on fedora systems (and not
just on opensuse32/64) [Thanks a.g.].
Fixed some of the testfiles. Clean-up/tidying the source code for some solvers that rely on
fortran solvers (resolved segfaults that occur when fortran code is expected to write run-time
messages).
Fixed glitches in gsl solvers for minimization.
Changes in 2.2.12.7
1. Functions fwrite/fseek/fread now work without glitches. 'fseek' requires file to be open prior
to seeking it. Code now passes the script 'test.r'.
Changes in 2.2.12.6
1. Fixed some memory leaks in read/write functions to the serial port. Added more options to the
open-function.
2. Added some more solvers: harminterp. Modified some existing solvers: linterp and findroot.
3. Reviewed sections of the code. Fixed some small memory leaks.
Changes in 2.2.12.5
1. Succesfully tested sockets on ethernet connected instruments - they have reached near-production quality.
Same minor changes in some system functions: tic()/toc() now have a microsecond precision.
Modifified/expanded some solvers, e.g., linterp.
Changes in 2.2.12.4
1. configuration of rlab is changed. The script ./configure no longer checks for the presence
of flex. Instead it uses the file provided with the distribution that was built with flex-2.5.35,
and then fixed by hand.
Added is option 'make scanner' that checks the file 'scanner.l' against local installation of flex
and tries to compile the result. Such installation, however, has to be put in 'Makefile.in' by hand
in line where '@FLEX@' is defined.
2. Added are few solvers, and some bugs and lack of features are corrected.
3. Sprannlib has been removed.
Changes in 2.2.12.3
1. The directory 'examples' transferred to the rlab-test collection of files that user has to
install in his/hers home directory, under ~/rlab/test.
2. More of the gsl: cumulative probability functions
Changes in 2.2.12.2
1. Changes in internal handling of open HDF5 files that removed few remaining glitches in basic
HDF5 handling functionality. The reliability is now at the production level for files
containing rlab recognized data.
2. Added more options to some existing solvers, e.g., findmins.
3. More functions and operators are now matrix optimized (addition, subtraction,
multiplication and division by element)
4. Added some gcc "#define"'s so that rlab builds properly on 32- and 64-bit
architectures. Fixed install script (was for version 2.2.11).
Changes in 2.2.12.1
Slight Modification of HDF5:
1. rlab scalar variables, i.e., matrices of size 1x1 are written as scalars (HDF5 data set
of rank 0) rather then as matrices (HDF5 data set of rank 2, with dimensions 1 and 1).
2. When writing non-scalar vectors and matrices to HDF5 objects, an attribute "Transpose"
with the value of unity is appended to them. This is to indicate to other software that
is used to access the data that it needs to be transposed.
If rlab reads a matrix that does not have that attribute with that value then it internally
transposes it.
3. The format of writing (and reading) of the sparse real and complex matrices to a HDF5 file
has changed to minimize recopying of the structure. An HDF5 storage of a sparse matrix is
a compound of 7 variable length arrays (6 integer, 1 double, or double-double - complex, that is).
Changes in 2.2.12.0
1. Read/write operations:
(a) Limited access to HDF5 - can read and write all data types native to rlab, and
can access atomic data sets within any HDF5 file. No support planned for special
HDF5 data types, e.g., compounds & c, except in the context of native data types.
(b) Integrated with cURL, so rlab can now read, download and upload web pages and files
to and from local filesystem and into the memory.
(c) Limited support for sockets, and their read and write operations.
2. More code cleaning and more glitches removed.
