dseries issueshttps://git.dynare.org/Dynare/dseries/issues2019-09-12T10:32:32Zhttps://git.dynare.org/Dynare/dseries/issues/41Logic for detection of bitwidth (32 or 64) for x13as binary is incorrect2019-09-12T10:32:32ZSébastien VillemotLogic for detection of bitwidth (32 or 64) for x13as binary is incorrectThe function `src/utilities/is/is64bit.m` is used to decide whether we use a 32-bit or 64-bit binary of x13as.
The logic of this test is doubly wrong.
First, it claims to check whether MATLAB or Octave is 32-bit or 64-bit, but it does it by checking the maximum size of matrix indices (there are 64-bit version of Octave which have a 32-bit index).
But, fundamentally, testing the bitwidth of MATLAB/Octave is not what we want, since x13as is an independent executable. We should rather test whether the operating system is 32- or 64-bit, as is currently done for the preprocessor binary.The function `src/utilities/is/is64bit.m` is used to decide whether we use a 32-bit or 64-bit binary of x13as.
The logic of this test is doubly wrong.
First, it claims to check whether MATLAB or Octave is 32-bit or 64-bit, but it does it by checking the maximum size of matrix indices (there are 64-bit version of Octave which have a 32-bit index).
But, fundamentally, testing the bitwidth of MATLAB/Octave is not what we want, since x13as is an independent executable. We should rather test whether the operating system is 32- or 64-bit, as is currently done for the preprocessor binary.https://git.dynare.org/Dynare/dseries/issues/11create methods to calculate T/T.A, G.A, A/A2018-09-21T04:38:15ZHoutan Bastanicreate methods to calculate T/T.A, G.A, A/AUseful transformations
Useful transformations
FerhatMihoubiFerhatMihoubihttps://git.dynare.org/Dynare/dseries/issues/10Problems when used under parfor2018-12-24T10:05:17ZTom HoldenProblems when used under parforSuppose you estimate a DSGE model by ML, forcing the optimisation route to use the parallel pool for derivative computations, then you get similar warnings to the ones mentioned here: https://www.mathworks.com/matlabcentral/newsreader/view_thread/251091 followed by an error.
Perhaps loadobj and saveobj methods are needed? At the moment I'm trying to hack in a solution by getting dsge_likelihood.m to convert structs to dseries if needed.
Suppose you estimate a DSGE model by ML, forcing the optimisation route to use the parallel pool for derivative computations, then you get similar warnings to the ones mentioned here: https://www.mathworks.com/matlabcentral/newsreader/view_thread/251091 followed by an error.
Perhaps loadobj and saveobj methods are needed? At the moment I'm trying to hack in a solution by getting dsge_likelihood.m to convert structs to dseries if needed.
https://git.dynare.org/Dynare/dseries/issues/8add new field, keep name field constant2018-09-21T04:38:15ZHoutan Bastaniadd new field, keep name field constantAs per the discussion in #7, add a new field to `dseries` that keeps track of the operations that have been taken on that series. In doing so, the `name` field should now remain constant.
As per the discussion in #7, add a new field to `dseries` that keeps track of the operations that have been taken on that series. In doing so, the `name` field should now remain constant.
Stéphane Adjemianstepan@dynare.orgStéphane Adjemianstepan@dynare.orghttps://git.dynare.org/Dynare/dseries/issues/4create syntax to compare dseries to numbers2018-12-24T09:19:50ZHoutan Bastanicreate syntax to compare dseries to numberse.g. if `a` is a dseries, the command `a>0` would return a vector/matrix of all elements in a.data that are greater than 0
e.g. if `a` is a dseries, the command `a>0` would return a vector/matrix of all elements in a.data that are greater than 0
https://git.dynare.org/Dynare/dseries/issues/3recreate repository with history from dynare.git2018-09-21T04:38:15ZHoutan Bastanirecreate repository with history from dynare.gitUse the `git subtree split` command to extract all commits that impacted this repository from within DynareTeam/dynare and recreate this repository with that commit history. As it currently stands, if we want to see the modifications made to a given file (as I just had to do for `load_csv_file_data.m` in #2), we have to go DynareTeam/dynare, `reset` to a point in time when the file existed in the repository, then use `gitk`.
Use the `git subtree split` command to extract all commits that impacted this repository from within DynareTeam/dynare and recreate this repository with that commit history. As it currently stands, if we want to see the modifications made to a given file (as I just had to do for `load_csv_file_data.m` in #2), we have to go DynareTeam/dynare, `reset` to a point in time when the file existed in the repository, then use `gitk`.
https://git.dynare.org/Dynare/dseries/issues/2CSV loading broken under Octave2019-09-12T10:26:47ZSébastien VillemotCSV loading broken under OctaveCommit DynareTeam/dynare@a2bd920d6f098b6a80be1f75d60cbd22c42899f5 (which has been incorporated into deseries with the location change of `load_csv_file_data`) has broken CSV loading under Octave.
There are two problems:
- if the first line is less than 4097 lines, then the io package is loaded but not used, so the data are not read
- in the other case, the code which is run is broken. In particular, the variables `withnames` and `withtime` are not defined (this bug is triggered by test number 4 of `name2tex.m`)
Moreover, I think it would make sense to allow people who don't have the io package to always use the replacement code (when fixed). Currently, people who don't have the io package can read files where the first line is > 4097, but not the smaller ones.
Commit DynareTeam/dynare@a2bd920d6f098b6a80be1f75d60cbd22c42899f5 (which has been incorporated into deseries with the location change of `load_csv_file_data`) has broken CSV loading under Octave.
There are two problems:
- if the first line is less than 4097 lines, then the io package is loaded but not used, so the data are not read
- in the other case, the code which is run is broken. In particular, the variables `withnames` and `withtime` are not defined (this bug is triggered by test number 4 of `name2tex.m`)
Moreover, I think it would make sense to allow people who don't have the io package to always use the replacement code (when fixed). Currently, people who don't have the io package can read files where the first line is > 4097, but not the smaller ones.
Sébastien VillemotSébastien Villemot