| 6 | | * Improvement of performance of internal TOL symbols searching methods and structures: |
| 7 | | o Syntax tree tokens detects constant literals (numbers, strings and dates) |
| 8 | | o Syntax tree tokens store references to undeletable system objects (type names, built-in functions, ...) |
| 9 | | o New local stack structure with a super fast dictionary tree with cost O(1) in RAM and CPU |
| 10 | | * Allowing use of a local Struct. |
| 11 | | See examples in test of bug 246 |
| 12 | | * Integrity of system constants: no longer the constant system objects will be modified. When an assignation of a constant is a done, an internal copy of the object is created so no unexpected collateral effect will take place. |
| 13 | | See examples in Tol tests (under construction) |
| 14 | | * By value pass of objects for builder operator (=), but on type Set. |
| 15 | | See examples in TOL tests CreationAndCopy |
| 16 | | * User Tol objects now have Description and Expression like separated functions, but only the description is modifiable by the user whereas the expression will only fill up Tol when the variable Real DoUnparseNodes is certain. When there is no description, user will receive the expression if exists, or the empty string. The Identify function returns the first non empty text of the list Name(), LocalName(), Expression(), Description() |
| | 6 | 1. Improvement of performance of internal TOL symbols searching methods and structures: |
| | 7 | 1. Syntax tree tokens detects constant literals (numbers, strings and dates) |
| | 8 | 1. Syntax tree tokens store references to undeletable system objects (type names, built-in functions, ...) |
| | 9 | 1. New local stack structure with a super fast dictionary tree with cost O(1) in RAM and CPU |
| | 10 | 1. Allowing use of a local Struct. See examples in test of bug 246 |
| | 11 | 1. Integrity of system constants: no longer the constant system objects will be modified. When an assignation of a constant is a done, an internal copy of the object is created so no unexpected collateral effect will take place. See examples in Tol tests (under construction) |
| | 12 | 1. By value pass of objects for builder operator (=), but on type Set. See examples in TOL tests CreationAndCopy |
| | 13 | 1. User Tol objects now have Description and Expression like separated functions, but only the description is modifiable by the user whereas the expression will only fill up Tol when the variable Real DoUnparseNodes is certain. When there is no description, user will receive the expression if exists, or the empty string. The Identify function returns the first non empty text of the list Name(), LocalName(), Expression(), Description() |
| 30 | | * Enable extended type messages in Write and WriteLn with an optional argument and new user functions Error and Warning in InitLibrary. |
| 31 | | * Best verbosity/mute handling in command line Tol Console. |
| 32 | | To know more about this feature write in your console |
| 33 | | tol -help |
| 34 | | * Functions that returns Anything can returns no object without send any error message and can be used in any scope. |
| 35 | | * Standard TOL log file start .log is moved to %APPDATA%\to\syslog in windows and ~/.tol in linux |
| 36 | | * New function PutLanguage to set the current human language of TOL messages |
| 37 | | * Each message of error or warning will adjoin a correlative number so that the user can find more easily first of them and follow the track more comfortably to them. |
| 38 | | * The Real variables TOL NError and Real NWarning have been created in order to return the number of errors and warnings emitted until the moment in the session, so that the users can verify in run time if a process is working correctly. |
| | 25 | 1. Enable extended type messages in Write and WriteLn with an optional argument and new user functions Error and Warning in InitLibrary. |
| | 26 | 1. Best verbosity/mute handling in command line Tol Console. To know more about this feature write in your console tol -help |
| | 27 | 1. Functions that returns Anything can returns no object without send any error message and can be used in any scope. |
| | 28 | 1. Standard TOL log file start .log is moved to %APPDATA%\to\syslog in windows and ~/.tol in linux |
| | 29 | 1. New function PutLanguage to set the current human language of TOL messages |
| | 30 | 1. Each message of error or warning will adjoin a correlative number so that the user can find more easily first of them and follow the track more comfortably to them. |
| | 31 | 1. The Real variables TOL NError and Real NWarning have been created in order to return the number of errors and warnings emitted until the moment in the session, so that the users can verify in run time if a process is working correctly. |
| 44 | | * The built-in TOL functions of Matrix type f01, F01 and Inv01 implement the density, the distribution and its inverse for variates with uncorrelated standard multinormal distribution. |
| 45 | | * The operators of matrices (+, -, *) now admit like second element as much a matrix as a real number. |
| | 37 | 1. The built-in TOL functions of Matrix type f01, F01 and Inv01 implement the density, the distribution and its inverse for variates with uncorrelated standard multinormal distribution. |
| | 38 | 1. The operators of matrices (+, -, *) now admit like second element as much a matrix as a real number. |
| 54 | | * Almost all bugs reported about TimeSet has been solved. |
| 55 | | * Faster methods for almost all successor an predecessor methods. |
| 56 | | * Enhanced cache methods. |
| 57 | | * Auto-abortable calculations in empty and bounded TimeSet's to avoid infinite cycles (a warning will be sent). |
| 58 | | * The begin is set to 1582-12-31 (Gregorian calendar begins at 1582-10-15) |
| | 47 | 1. Almost all bugs reported about TimeSet has been solved. |
| | 48 | 1. Faster methods for almost all successor an predecessor methods. |
| | 49 | 1. Enhanced cache methods. |
| | 50 | 1. Auto-abortable calculations in empty and bounded TimeSet's to avoid infinite cycles (a warning will be sent). |
| | 51 | 1. The begin is set to 1582-12-31 (Gregorian calendar begins at 1582-10-15) |
| 73 | | * Enhancement of unknown treatment in distribution functions. |
| 74 | | * New faster methods for truncated normal distribution: |
| 75 | | o Scalar: DensMultiNormalTrunc, DensTruncatedNormal, DistTruncatedNormal, DistTruncatedNormalInv, DrawTruncatedNormal |
| 76 | | o Vectorial: RandTruncatedMultNormal, LogDensTruncatedMultNormal |
| 77 | | * New methods Markov Chain Monte Carlo (MCMC) for the sampling of implicitly defined probability distributions using built-in functions GibbsSampler and MetropolisHastings. |
| 78 | | To know more about this feature write in your terminal screen: |
| 79 | | tol -vU -c"WriteLn(Description(GibbsSampler))" \ |
| 80 | | -c"WriteLn(Description(MetropolisHastings))" |
| 81 | | * Solving bugs in definition, naming and using of R2, Pearson, Correlation and other statistics |
| 82 | | * New package of TOL built-in functions to allow TOL users to work with GSL functions which both returning and arguments are all of numeric type. |
| 83 | | Read more here |
| 84 | | * Enhancements in function Estimate (Max. Likelihood ARIMA models estimator): |
| 85 | | o When a step gives worst result than previous iteration estimation of missing values must remains unchanged. |
| 86 | | o An error will be emited if an input hasn't InputDef structure. |
| 87 | | o In the ARMAAutoCovarianzeVector function used internally by Estimate there was an invalid access of memory when there is no part AR in a model and a smaller number of autocovarianzas is asked for or just as degree MA, which could give rise to falls or strange behaviors of random and irreproducible form. |
| | 66 | 1. Enhancement of unknown treatment in distribution functions. |
| | 67 | 1. New faster methods for truncated normal distribution: |
| | 68 | 1. Scalar: DensMultiNormalTrunc, DensTruncatedNormal, DistTruncatedNormal, DistTruncatedNormalInv, DrawTruncatedNormal |
| | 69 | 1. Vectorial: RandTruncatedMultNormal, LogDensTruncatedMultNormal |
| | 70 | 1. New methods Markov Chain Monte Carlo (MCMC) for the sampling of implicitly defined probability distributions using built-in functions GibbsSampler and MetropolisHastings. To know more about this feature write in your terminal screen: tol -vU -c"WriteLn(Description(GibbsSampler))" \ -c"WriteLn(Description(MetropolisHastings))" |
| | 71 | 1. Solving bugs in definition, naming and using of R2, Pearson, Correlation and other statistics |
| | 72 | 1. New package of TOL built-in functions to allow TOL users to work with GSL functions which both returning and arguments are all of numeric type. |
| | 73 | 1. Enhancements in function Estimate (Max. Likelihood ARIMA models estimator): |
| | 74 | 1. When a step gives worst result than previous iteration estimation of missing values must remains unchanged. |
| | 75 | 1. An error will be emited if an input hasn't InputDef structure. |
| | 76 | 1. In the ARMAAutoCovarianzeVector function used internally by Estimate there was an invalid access of memory when there is no part AR in a model and a smaller number of autocovarianzas is asked for or just as degree MA, which could give rise to falls or strange behaviors of random and irreproducible form. |
