3/7/2023 0 Comments Activetcl download 8.5.18![]() PSF cautions Mac users not to use tcl 8.6 because current Pythons 2.7 series and Python 3.4 won't link to it. On the offending Mac, I was using ActiveState TCL 8.5.18. I noted that it was using ActiveState TCL 8.5.15. Tipoff 2: I had access to another Mac on which everything worked perfectly between PSF (Python Software Foundation) Python and Julia. This suggests that the handshaking between Julia and Python was just fine, but the handshaking between Python and tcl/tk was suspect. After close(), the fignum list was empty, but the plot window was still there on the screen for all the world to see (well, maybe just me.). Tipoff 1: After creating the first plot, there was a fignum returned for it. In the repl, PyObjects that are returned seem OK (at least they appear to be similar to later PyObjects for subsequent figures). This is true even when a figure IS explicitly requested and assigned to a variable in Julia, e.g: There is something about how the first figure object is generated that is getting lost so that a subsequent close() call back to matplotlib is not closing the "current" figure-when we are doing the first plot. I tried but that is not really meant to work across "using" I think. I am not failing on those tests.Īny suggestions for how to debug the calls to python and returns from python through the PyPlot functions? The preprocessor steps ( that generate all of the Julia callables don't make it easy to insert print statements. Looked through source and the primary differences for conda are the tests for pre-reqs and the installation of missing pre-reqs. Using lower level direct call seem to work in both cases so P圜all/PyPlot are certainly making the connection to the installed Python. Outputs to the repl are essentially the same. Looked at pycall deps: could see no major differences except the paths that were set and true/false for conda =. Conda python closes the first plot system python does not. Tried both the self-contained Python (conda) and system Python. Uninstalled/reinstalled Julia and all packages. Spent 5 hours on this over the past 2 days. What tests can I run? Are there are other parts of matplotlib I can call to get info? I have used plt :get_fignums and it returns 0-element Array-though it is there for all to see. I really will help here, but I am out of things to look at. There is something about how PyPlot/P圜all calls through to Python that orphans the first plot window. But, I really, really don't want to do that and should have to. The only thing that works is to use the conda Python. But, this doesn't work with either matplotlib 1.4.3 or 1.5.0. Note that the second plot window created by Julia always behaves normally.įurther, I downgraded matplotlib from 1.5.0 to 1.4.3 with the same problem.įurther, I downgraded from Python 2.7.10 to 2.7.9, which was the last version I remember working. When I run the same instance of Python (2.7.10 from PSF) through Ipython and create a trivial plot, I am able to close the plot window. Just confirming that this appears to be a Julia related problem. Not closing the plot remains a serious issue. On the mac managing Python is really easy. ![]() I am not sure that reinstalling os x will fix this. So, now I must use conda and perhaps never have system pythons installed properly. There is no reliable and safe way to purge usr/bin without destroying something. A pox on Unix and all of its derivatives. This means I have a mangled usr/bin and usr/local/bin. It still doesn't work even with matplotlib 1.4.3. Now, I'll go back a version and try to confirm this. There is also something about the latest python 2.7.10 installer that even with matplotlib 1.4.3 the first figure window drawn cannot be closed. You can use it, of course, if you path to it properly. By the way, there is nothing wrong with P圜all managing the dependency stack "internally" with conda. I will confirm this with standalone Python. The problem above appears to be caused by using matplotlib 1.5.0. This brings in Python 2.7.10 and matplotlib 1.4.3. Looking at the Python versions I was chasing the wrong theory.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |