Limits of `Polynomials.PolyCompat`?

B-LIE commented 4 years ago

In the past (i.e., some months ago), functions polyfit, polyval, etc. worked. I have the impression that polyfit has been replaced by fit, and that polyval has been eliminated.

I have tried to use commands:

using Polynomials
Polynomials.PolyCompat

in the hope that this would make my old code work, but I get error messages.

Should this work?

jverzani commented 4 years ago

Yes, but I think you are missing a "using Polynomials.PolyCompat" (the using isn't in your message). If it fails after that, please do show an example as the goal was to support the legacy code just by adding that one "using" command.

B-LIE commented 4 years ago

Ah. OK…

Question: the new fit function doesn’t allow for a variable symbol as argument, right? I.e.

fit(x,y,3,:s)

doesn’t work. Right?

Is the trick to do:

fit(x,y,3) |> coeffs |> x -> Polynomial(x,:s)

?

-B

From: john verzani notifications@github.com Sent: Wednesday, July 29, 2020 18:50 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Yes, but I think you are missing a "using Polynomials.PolyCompat" (the using isn't in your message). If it fails after that, please do show an example as the goal was to support the legacy code just by adding that one "using" command.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-665778295&data=02%7C01%7C%7C8552e24349cb47cc888108d833df5f92%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316381750338072&sdata=ZUFwnHMWmZjM7rTJsmIaxjON16deVWysCAR5%2Fxs9Ynk%3D&reserved=0, or unsubscribehttps://eur06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM74T4HV66Z772DRTARTR6BHJ3ANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C8552e24349cb47cc888108d833df5f92%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316381750338072&sdata=wBB%2F4%2BeaV0GV42shtevsprO7YPMPYYPGTW%2BIeO4z%2Btw%3D&reserved=0.

jverzani commented 4 years ago

It is a keyword now. Try adding fit(x,y,3,var=:s).

B-LIE commented 4 years ago

Great! Thanks!

And polyval has been removed in the new system?

-B

From: john verzani notifications@github.com Sent: Wednesday, July 29, 2020 18:56 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

It is a keyword now. Try adding fit(x,y,3,var=:s).

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-665781681&data=02%7C01%7C%7C5bc0ad816c5948b04b6b08d833e03dd9%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316385478642520&sdata=e2VXD%2BKz9T%2FQQZo9Cosn6dwlhsyFRd4SGsmGo6WIUeY%3D&reserved=0, or unsubscribehttps://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM72GEVCS343YMPHOANLR6BIBFANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C5bc0ad816c5948b04b6b08d833e03dd9%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316385478652519&sdata=%2BpwHKhiffg%2F3TQwVqW4lDrpqR4oB0Yq9whuhCTweoKQ%3D&reserved=0.

jverzani commented 4 years ago

Yes, just do p(x) instead. (There is a non exported evalpoly that might be useful for broadcasting purposes).

B-LIE commented 4 years ago

Great! Thanks – and sorry for dropping out after early April 2020 – I’ve been busy. I’ll try to play with Polynomials.jl in the near future. (I had some questions in March/April about matrices of polynomials…)

BR, Bernt

From: john verzani notifications@github.com Sent: Wednesday, July 29, 2020 18:59 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Yes, just do p(x) instead. (There is a non exported evalpoly that might be useful for broadcasting purposes).

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur05.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-665783345&data=02%7C01%7C%7C4b930a89dc794b29126208d833e0af36%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316387381712722&sdata=18VA109qfK1VLFhAUxFyL7ErUX4oGhiXQY8w4kbj%2Baw%3D&reserved=0, or unsubscribehttps://eur05.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM73KZAWQ7LJ42RJKQ6TR6BINBANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C4b930a89dc794b29126208d833e0af36%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316387381712722&sdata=i96oFqiQAqmU98W7GC%2FNwo9nRFwhQPuZ%2BpADQUqwvXM%3D&reserved=0.

jverzani commented 4 years ago

Any feedback when you get a chance would be great. If you are interested, I have a PR that looked at allowing matrices as coefficients. It proves mostly successful when the size of the matrix is in the type information (e.g. SArrays) and less so when not. Anyways, I never merged it, as I didn't get any feedback on it and there are a few needed workarounds I recall being more hacky than I would prefer.

B-LIE commented 4 years ago

I’ll get back to some of the work I did in the early Lockdown period. I.e., I’ll look at it throughout the fall semester.

-B

From: john verzani notifications@github.com Sent: Wednesday, July 29, 2020 19:19 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Any feedback when you get a chance would be great. If you are interested, I have a PR that looked at allowing matrices as coefficients. It proves mostly successful when the size of the matrix is in the type information (e.g. SArrays) and less so when not. Anyways, I never merged it, as I didn't get any feedback on it and there are a few needed workarounds I recall being more hacky than I would prefer.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-665793024&data=02%7C01%7C%7C241720ea266240ef8a6408d833e37421%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316399275870286&sdata=3yHP8mBVw%2FrRgrevW8vLobJszHzLS72P6HMQ%2FMYP2SU%3D&reserved=0, or unsubscribehttps://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM74GEWS3463SZTCOIQTR6BKXNANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C241720ea266240ef8a6408d833e37421%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316399275880282&sdata=VzTNzp5a30R1kpGnJ7DeebxaU1NqeIIXFnzd%2B6nIWbs%3D&reserved=0.

B-LIE commented 4 years ago

I occasionally get a mysterious error message when running fit: [cid:image001.png@01D66660.1BDAB560]

The weird thing is that if I run it a second (or third, …) time, the error message disappears, and it works.

Have you seen such a thing before? NOTE: I run the code in Jupyterlab, and I’m not sure how solid Jupyterlab is… (I’ve seen other cases where code doesn’t work properly in Jupyterlab, while it works in Juno or VScode.)

-Bernt

From: john verzani notifications@github.com Sent: Wednesday, July 29, 2020 19:19 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Any feedback when you get a chance would be great. If you are interested, I have a PR that looked at allowing matrices as coefficients. It proves mostly successful when the size of the matrix is in the type information (e.g. SArrays) and less so when not. Anyways, I never merged it, as I didn't get any feedback on it and there are a few needed workarounds I recall being more hacky than I would prefer.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-665793024&data=02%7C01%7C%7C241720ea266240ef8a6408d833e37421%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316399275870286&sdata=3yHP8mBVw%2FrRgrevW8vLobJszHzLS72P6HMQ%2FMYP2SU%3D&reserved=0, or unsubscribehttps://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM74GEWS3463SZTCOIQTR6BKXNANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C241720ea266240ef8a6408d833e37421%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316399275880282&sdata=VzTNzp5a30R1kpGnJ7DeebxaU1NqeIIXFnzd%2B6nIWbs%3D&reserved=0.

B-LIE commented 4 years ago

Hm… I moved the code over to VScode, and I get the same error message there.

If you are interested in checking it out, see attached Jupyter notebook + 2 CSV files of data. NOTE: you need to change two paths in cell 2.

-B

From: Bernt Lie Bernt.Lie@hotmail.com Sent: Thursday, July 30, 2020 10:56 To: JuliaMath/Polynomials.jl reply@reply.github.com; JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: Author author@noreply.github.com Subject: RE: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

I occasionally get a mysterious error message when running fit: [cid:image001.png@01D66669.864A3DE0]

The weird thing is that if I run it a second (or third, …) time, the error message disappears, and the code works.

Have you seen such a thing before? NOTE: I run the code in Jupyterlab, and I’m not sure how solid Jupyterlab is… (I’ve seen other cases where code doesn’t work properly in Jupyterlab, while it works in Juno or VScode.)

-Bernt

From: john verzani notifications@github.com<mailto:notifications@github.com> Sent: Wednesday, July 29, 2020 19:19 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com<mailto:Polynomials.jl@noreply.github.com> Cc: B-LIE bernt.lie@hotmail.com<mailto:bernt.lie@hotmail.com>; Author author@noreply.github.com<mailto:author@noreply.github.com> Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Any feedback when you get a chance would be great. If you are interested, I have a PR that looked at allowing matrices as coefficients. It proves mostly successful when the size of the matrix is in the type information (e.g. SArrays) and less so when not. Anyways, I never merged it, as I didn't get any feedback on it and there are a few needed workarounds I recall being more hacky than I would prefer.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-665793024&data=02%7C01%7C%7C241720ea266240ef8a6408d833e37421%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316399275870286&sdata=3yHP8mBVw%2FrRgrevW8vLobJszHzLS72P6HMQ%2FMYP2SU%3D&reserved=0, or unsubscribehttps://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM74GEWS3463SZTCOIQTR6BKXNANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C241720ea266240ef8a6408d833e37421%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637316399275880282&sdata=VzTNzp5a30R1kpGnJ7DeebxaU1NqeIIXFnzd%2B6nIWbs%3D&reserved=0.

jverzani commented 4 years ago

I can't find the data. There was recently a different, but perhaps related issue where the user identified a graphing package (pyplot) as causing a conflict. To be honest, I can't see the connection, but... Anyways, I'll have a look if I can figure out where the numbers are.

B-LIE commented 4 years ago

File WatetSaturated.csv. You need to put it in a directory corresponding to the path in cell 2.

Get Outlook for Androidhttps://aka.ms/ghei36

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 12:52:10 PM To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

I can't find the data. There was recently a different, but perhaps related issue where the user identified a graphing package (pyplot) as causing a conflict. To be honest, I can't see the connection, but... Anyways, I'll have a look if I can figure out where the numbers are.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666294875&data=02%7C01%7C%7C7d29b38a6aa04139ed1f08d834769c56%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317031310427026&sdata=032O8PxRim2PGDXDMlen390BN7tTZdQbtNjyASYv9g8%3D&reserved=0, or unsubscribehttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM775KYH7Q2JKZDFEAU3R6FGFVANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C7d29b38a6aa04139ed1f08d834769c56%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317031310437019&sdata=AjpRToNaHJcSh5tFV9XM7dAHGkznqDGyGhwkU5VIvHo%3D&reserved=0.

jverzani commented 4 years ago

Where do I find the file? Does it get attached to this issue somehow?

B-LIE commented 4 years ago

Hm... Did you get the Jupiter notebook? I attached both the Jupyter notebook + 2 csv-files in the same e-mail. Maybe the system strips off attachments?

Get Outlook for Androidhttps://aka.ms/ghei36

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 1:28:20 PM To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Where do I find the file? Does it get attached to this issue somehow?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666309780&data=02%7C01%7C%7Ce83d848c09414f9274de08d8347baad9%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317053027638101&sdata=61jVGmYdyXYv%2BsKTIc6FjhSbxlj0yny2%2Fx%2BYgiN%2FYms%3D&reserved=0, or unsubscribehttps://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM7YNFUQR7ZBCQTDESH3R6FKNJANCNFSM4PLYPZ4Q&data=02%7C01%7C%7Ce83d848c09414f9274de08d8347baad9%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317053027648097&sdata=D%2FKwSnHLz%2FW3ZCbRM8B%2F16hHkEdUFI%2F3kElhh2gu03g%3D&reserved=0.

jverzani commented 4 years ago

I think it strips it off. Try sending to jverzani at gmail

On Thu, Jul 30, 2020 at 7:31 AM B-LIE notifications@github.com wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edu

B-LIE commented 4 years ago

Hm... The Jupyter notebook I sent you works on my laptop (where I have installed Julia v. 1.4.0). The problem appears on my work desktop (with Julia v. 1.4.2).

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 13:57 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

I think it strips it off. Try sending to jverzani at gmail

On Thu, Jul 30, 2020 at 7:31 AM B-LIE notifications@github.com<mailto:notifications@github.com> wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edumailto:verzani@math.csi.cuny.edu

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666321123&data=02%7C01%7C%7Ca1850af697f645bade5d08d8347fb223%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317070335792805&sdata=jmnxRedzKU6LvLgfOcnrOTjjebWhIXMGKUMiJstJAa4%3D&reserved=0, or unsubscribehttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM7YXEPXDYZDYMZCVEHLR6FNZRANCNFSM4PLYPZ4Q&data=02%7C01%7C%7Ca1850af697f645bade5d08d8347fb223%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317070335802799&sdata=g5yijoMbVO08K0QDT9A9coOHfNsyhP9cCfxeZ07NN0Y%3D&reserved=0.

jverzani commented 4 years ago

That worked, but unfortunately I can't replicate the issue. fit(x,y,3) basically does this:

A = vander(Polynomial, a, 3)
pinv(A) * y

It could also be A \ y. I was expecting A to be poorly conditioned, but it doesn't appear to be.

If you bump into the error again, could you try this directly and see if the same error occurs?

B-LIE commented 4 years ago

OK – I’ll test it.

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 14:20 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

That worked, but unfortunately I can't replicate the issue. fit(x,y,3) basically does this:

A = vander(Polynomial, a, 3)

pinv(A) * y

It could also be A \ y. I was expecting A to be poorly conditioned, but it doesn't appear to be.

If you bump into the error again, could you try this directly and see if the same error occurs?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666330573&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960989984&sdata=DQD24uGnf1Ex3wXQMcWE2qUWz6dF7V%2FJrrqKod2D0Xg%3D&reserved=0, or unsubscribehttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75DFAILNSB6HZSCB4LR6FQOVANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960999980&sdata=9U9F1HHwh4EgS39cKGt9ou10swFZysTjkaQQO1giWxE%3D&reserved=0.

B-LIE commented 4 years ago

Hm... it is a little unclear to me how vander works: [cid:image001.png@01D6667D.B3E80200]

[minor typo: “psuedo” should be “pseudo”]

What is Polynomial? Should it be Polynomial (verbatim), or the name of a polynomial? I haven’t created a polynomial yet. And what is a? Is it the input, i.e., x?

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 14:20 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

That worked, but unfortunately I can't replicate the issue. fit(x,y,3) basically does this:

A = vander(Polynomial, a, 3)

pinv(A) * y

It could also be A \ y. I was expecting A to be poorly conditioned, but it doesn't appear to be.

If you bump into the error again, could you try this directly and see if the same error occurs?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666330573&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960989984&sdata=DQD24uGnf1Ex3wXQMcWE2qUWz6dF7V%2FJrrqKod2D0Xg%3D&reserved=0, or unsubscribehttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75DFAILNSB6HZSCB4LR6FQOVANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960999980&sdata=9U9F1HHwh4EgS39cKGt9ou10swFZysTjkaQQO1giWxE%3D&reserved=0.

B-LIE commented 4 years ago

OK – figured it out. [cid:image001.png@01D6667E.616D3DF0]

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 14:20 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

That worked, but unfortunately I can't replicate the issue. fit(x,y,3) basically does this:

A = vander(Polynomial, a, 3)

pinv(A) * y

It could also be A \ y. I was expecting A to be poorly conditioned, but it doesn't appear to be.

If you bump into the error again, could you try this directly and see if the same error occurs?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666330573&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960989984&sdata=DQD24uGnf1Ex3wXQMcWE2qUWz6dF7V%2FJrrqKod2D0Xg%3D&reserved=0, or unsubscribehttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75DFAILNSB6HZSCB4LR6FQOVANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960999980&sdata=9U9F1HHwh4EgS39cKGt9ou10swFZysTjkaQQO1giWxE%3D&reserved=0.

B-LIE commented 4 years ago

OK – as I mentioned, there is no problem on my laptop (Julia v. 1.4.0). However, on my desktop (Julia v. 1.4.2) there is a problem with statement:

Julia> k_pol = fit(Tw_s,k_Lw/k_Lw0,2;var=Symbol("(ΔT₀/T₀)"))

When I run it the first time, I get an error message related to LAPACKException(2) or LAPACKException(4)

If I rerun the same cell, the error message disappears. If I run it again, it returns, etc.

It is somewhat mysterious that there is a difference in behavior on my desktop and on my laptop. This may have to do with (i) Julia version number, (ii) Polynomials version number, (iii) Plots version number, (iv) PyPlot version number, (v) hardware configuration, etc. I doubt that the problem is related to IJulia and JupyterLab, simply because the same problem appears when I run the code in VScode.

--

I have tried the following:

If I replace the PyPlot backend of Plots with GR, then the problem disappears!! So there appears to be a problem with the PyPlot backend.
If I replace “fit” with “Polynomials.fit”, this has no effect. In other words: the problem is not due to a namespace conflict with PyPlot or JupyterLab.

This seems to reduce the cause to something related to PyPlot-Plots-hardware configuration.

--

NOTE: I also tested this on my old desktop, with Julia v. 1.4.0. Both the old desktop and the new desktop use the same version no. of packages Polynomials, Plots, PyPlot, and IJulia. On the old desktop, (Julia v. 1.4.0), the code works. On the new desktop (Julia v. 1.4.2), the code fails sometimes. I’ll upgrade the old desktop to Julia v. 1.4.2…

So – why don’t I just switch from using PyPlot to using GR? Because PyPlot has much better support for LaTeX symbols (although not perfect!!). The current support for LaTeX in the GR backend of Plots is really, really lacking.

Fortunately, at JuliaCon 2020 yesterday, the people behind GR presented a vastly improved support for LaTeX. I do, however, suspect that it will take some time before they have upgraded the GR backend to Plots…

Thanks for help!

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 14:20 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

That worked, but unfortunately I can't replicate the issue. fit(x,y,3) basically does this:

A = vander(Polynomial, a, 3)

pinv(A) * y

It could also be A \ y. I was expecting A to be poorly conditioned, but it doesn't appear to be.

If you bump into the error again, could you try this directly and see if the same error occurs?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666330573&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960989984&sdata=DQD24uGnf1Ex3wXQMcWE2qUWz6dF7V%2FJrrqKod2D0Xg%3D&reserved=0, or unsubscribehttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75DFAILNSB6HZSCB4LR6FQOVANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960999980&sdata=9U9F1HHwh4EgS39cKGt9ou10swFZysTjkaQQO1giWxE%3D&reserved=0.

B-LIE commented 4 years ago

OK – I have updated my old desktop to Julia 1.4.2, and run exactly the same versions of Polynomials, Plots, PyPlot, and IJulia on the old and new desktops. On my old desktop, things work. On my new desktop, errors occur.

Hm… mysterious. Maybe some other packages are different.

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 14:20 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

That worked, but unfortunately I can't replicate the issue. fit(x,y,3) basically does this:

A = vander(Polynomial, a, 3)

pinv(A) * y

It could also be A \ y. I was expecting A to be poorly conditioned, but it doesn't appear to be.

If you bump into the error again, could you try this directly and see if the same error occurs?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666330573&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960989984&sdata=DQD24uGnf1Ex3wXQMcWE2qUWz6dF7V%2FJrrqKod2D0Xg%3D&reserved=0, or unsubscribehttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75DFAILNSB6HZSCB4LR6FQOVANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C6d9a089726bf4d14c1cb08d83482de98%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317083960999980&sdata=9U9F1HHwh4EgS39cKGt9ou10swFZysTjkaQQO1giWxE%3D&reserved=0.

jverzani commented 4 years ago

The other user said their similar issue was solved by installing MKL.jl. My suspicion is not that fit is different, but perhaps there is some oddity in what pinv calls. If I knew that, we could replace that call with the \ approach. Anyways, sorry about all the detective work.

B-LIE commented 4 years ago

I’m trying to install one-by-one additional packages on my old desktop that were lacking in the new one. I’ll report when I have tested the remaining 4 packages on the old desktop.

Installing MKL.jl – is that available in the repository? If I install it, do I have to do any other changes? (NOTE: I don’t have MKL.jl installed on the PCs where things work…)

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 16:39 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

The other user said their similar issue was solved by installing MKL.jl. My suspicion is not that fit is different, but perhaps there is some oddity in what pinv calls. If I knew that, we could replace that call with the \ approach. Anyways, sorry about all the detective work.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666406856&data=02%7C01%7C%7C29baa7e854f047d8a75808d83496439e%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317167261602619&sdata=lJcJgLxzgG1VxZKiTIWGjQmqHMlBXfPBwl8exybnevw%3D&reserved=0, or unsubscribehttps://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM7ZAMB45UMGSTWX5YFLR6GAXLANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C29baa7e854f047d8a75808d83496439e%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317167261612620&sdata=yNQN8WbLi0LpZfrZkSIqsYYEhlP7jZLNafd12TALNdM%3D&reserved=0.

jverzani commented 4 years ago

I don't know about MKL, but that is a poor solution. Better to see where the root cause is.

B-LIE commented 4 years ago

Hm. I installed MKL.jl. After that, both Jupyterlab and VScode crashed at the first instance of calling fit(). So I’ll remove it again.

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 17:23 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

I don't know about MKL, but that is a poor solution. Better to see where the root cause is.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666450834&data=02%7C01%7C%7C62d04762d4ec492ec9f408d8349c6b18%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317193694114338&sdata=JFAqop9E80njqyUDKYN8JDJ1uYn00PqcyMagBKPho5c%3D&reserved=0, or unsubscribehttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75NMII5HGMIY2A3UNTR6GF4RANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C62d04762d4ec492ec9f408d8349c6b18%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317193694114338&sdata=n%2FoCvTbrCvdahs4Sn4j1i0zh%2B8mB45%2B02Rn1OMq%2Bb%2FY%3D&reserved=0.

B-LIE commented 4 years ago

OK – due to the problem with MKL, I ended up uninstalling Julia completely as well as Python/miniconda, and reinstalling everything from scratch.

Unfortunately, the problem of “fit” persists.

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 17:23 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

I don't know about MKL, but that is a poor solution. Better to see where the root cause is.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666450834&data=02%7C01%7C%7C62d04762d4ec492ec9f408d8349c6b18%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317193694114338&sdata=JFAqop9E80njqyUDKYN8JDJ1uYn00PqcyMagBKPho5c%3D&reserved=0, or unsubscribehttps://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75NMII5HGMIY2A3UNTR6GF4RANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C62d04762d4ec492ec9f408d8349c6b18%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317193694114338&sdata=n%2FoCvTbrCvdahs4Sn4j1i0zh%2B8mB45%2B02Rn1OMq%2Bb%2FY%3D&reserved=0.

jverzani commented 4 years ago

If instead of fit(Polynomial, x, y, d) you use the backslash, as with:

A = vander(Polynomial, x, d)
Polynomial(A \ y; var=:s)

Do you have the same issue? (My suspicions center around pinv, but that is only a hunch.)

On Thu, Jul 30, 2020 at 11:59 AM B-LIE notifications@github.com wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edu

B-LIE commented 4 years ago

I’ll take a look tomorrow – I just went home for supper :-o. So I’ll write my own fit_test function using backslash instead of pinv :-).

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 18:16 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

If instead of fit(Polynomial, x, y, d) you use the backslash, as with:

A = vander(Polynomial, x, d)
Polynomial(A \ y; var=:s)

Do you have the same issue? (My suspicions center around pinv, but that is only a hunch.)

On Thu, Jul 30, 2020 at 11:59 AM B-LIE notifications@github.com<mailto:notifications@github.com> wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edumailto:verzani@math.csi.cuny.edu

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666499458&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=wd6SdXf6JX5djc6gn33zAQfmF25GQoCaeI5kMFcRGBg%3D&reserved=0, or unsubscribehttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75LX6X4SV6NEHVTFMTR6GMD7ANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=TcdCBCVerV%2BJBIKJPh3BU7oZbtP3BLMzhR3wNZok9p0%3D&reserved=0.

B-LIE commented 4 years ago

OK – some more testing – this is scary:

[cid:image001.png@01D66724.61F24800]

-Bernt

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 18:16 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

If instead of fit(Polynomial, x, y, d) you use the backslash, as with:

A = vander(Polynomial, x, d)
Polynomial(A \ y; var=:s)

Do you have the same issue? (My suspicions center around pinv, but that is only a hunch.)

On Thu, Jul 30, 2020 at 11:59 AM B-LIE notifications@github.com<mailto:notifications@github.com> wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edumailto:verzani@math.csi.cuny.edu

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666499458&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=wd6SdXf6JX5djc6gn33zAQfmF25GQoCaeI5kMFcRGBg%3D&reserved=0, or unsubscribehttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75LX6X4SV6NEHVTFMTR6GMD7ANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=TcdCBCVerV%2BJBIKJPh3BU7oZbtP3BLMzhR3wNZok9p0%3D&reserved=0.

B-LIE commented 4 years ago

More of scary testing: [cid:image001.png@01D66726.E46857F0]

… and if I reset the kernel, use the GR backend for Plots, and re-run the code: [cid:image002.png@01D66727.16CF5400]

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 18:16 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

If instead of fit(Polynomial, x, y, d) you use the backslash, as with:

A = vander(Polynomial, x, d)
Polynomial(A \ y; var=:s)

Do you have the same issue? (My suspicions center around pinv, but that is only a hunch.)

On Thu, Jul 30, 2020 at 11:59 AM B-LIE notifications@github.com<mailto:notifications@github.com> wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edumailto:verzani@math.csi.cuny.edu

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666499458&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=wd6SdXf6JX5djc6gn33zAQfmF25GQoCaeI5kMFcRGBg%3D&reserved=0, or unsubscribehttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75LX6X4SV6NEHVTFMTR6GMD7ANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=TcdCBCVerV%2BJBIKJPh3BU7oZbtP3BLMzhR3wNZok9p0%3D&reserved=0.

B-LIE commented 4 years ago

OK… some more results:

If I move all the polynomial fitting computations in front of the plotting, then the code works.
I suspect some symbols I use in the plotting with PyPlot interacts with the LinearAlgebra package.

Still scary…

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 18:16 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

If instead of fit(Polynomial, x, y, d) you use the backslash, as with:

A = vander(Polynomial, x, d)
Polynomial(A \ y; var=:s)

Do you have the same issue? (My suspicions center around pinv, but that is only a hunch.)

On Thu, Jul 30, 2020 at 11:59 AM B-LIE notifications@github.com<mailto:notifications@github.com> wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edumailto:verzani@math.csi.cuny.edu

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666499458&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=wd6SdXf6JX5djc6gn33zAQfmF25GQoCaeI5kMFcRGBg%3D&reserved=0, or unsubscribehttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75LX6X4SV6NEHVTFMTR6GMD7ANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=TcdCBCVerV%2BJBIKJPh3BU7oZbtP3BLMzhR3wNZok9p0%3D&reserved=0.

B-LIE commented 4 years ago

Argh. I’m giving up… Here is the situation:

My code consists of 2 sets of physical property data. The first set is for liquid water, the second for air at atmospheric pressure. There are more data for liquid water than for air.
When fitting and plotting physical properties for air (the second phase), there is no problem – everything runs smoothly, and all results are computed without error messages.
The problem is in the first phase with water data.
If I do all the fitting using Polynomials before I start plotting the water data, there is no problem – things run smoothly. The disadvantage of this is that it reduces the pedagogics/interactive structure of the presentation.
If I do the fitting for one property, then plots this, do the fitting for the second property, and plot this – the code crashes on fitting the second property. Essentially, Julia claims that the Vandermonde matrix for the second fitting has rank zero. HOWEVER, if I call the rank function a second time before I do the plotting, the rank is correctly computed as 5.
The “reasonable” conclusion is that the plotting command issued after the first fitting has messed up with the LinearAlgebra package when I call the rank function in the second fitting. [Why? Because if I do all the fitting prior the any plotting, there is no problem.]
So… some special symbol I use in the plotting command that messes up the LinearAlgebra package? Unfortunately not -- any plotting messes up the LinearAlgebra package apparently [I tested with “plot(sin)” without any other Plots command, and then the rank problem was introduced].
The problem only appears if I use the PyPlot backend. If I use the GR backend, then the problem disappears. [But the GR backend still misses a lot of symbols that I need…]
If I manually run the code blocks with fitting + plotting two times in a row, then the code works.
And… the problem only appears on my brand new HP desktop computer, not on my old Dell desktop (running exactly the same versions of the packages). And not on my laptop (running an older version of Julia).

It is frustrating to do debugging when there is no apparent reason for the problem. The code crashes in the first phase (water data), but not in the second phase (air data, which are fewer than the water data).

-B

From: john verzani notifications@github.com Sent: Thursday, July 30, 2020 18:16 To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

If instead of fit(Polynomial, x, y, d) you use the backslash, as with:

A = vander(Polynomial, x, d)
Polynomial(A \ y; var=:s)

Do you have the same issue? (My suspicions center around pinv, but that is only a hunch.)

On Thu, Jul 30, 2020 at 11:59 AM B-LIE notifications@github.com<mailto:notifications@github.com> wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edumailto:verzani@math.csi.cuny.edu

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-666499458&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=wd6SdXf6JX5djc6gn33zAQfmF25GQoCaeI5kMFcRGBg%3D&reserved=0, or unsubscribehttps://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM75LX6X4SV6NEHVTFMTR6GMD7ANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C06085db82a9c4ffc6b1c08d834a3d93f%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317225604690244&sdata=TcdCBCVerV%2BJBIKJPh3BU7oZbtP3BLMzhR3wNZok9p0%3D&reserved=0.

jverzani commented 4 years ago

Wow, that is subtle, and sadly consistent with the other report that was related. There a new gaming machine was the only machine that caused issues. I'll see if I can recreate locally somehow so I can slowly peel back the plot commands. There must be something triggering this.

B-LIE commented 4 years ago

My new desktop has a 10th generation (?) i9 processor. I'll check what type of graphics card there is.

B

Get Outlook for Androidhttps://aka.ms/ghei36

From: john verzani notifications@github.com Sent: Friday, July 31, 2020 1:13:20 PM To: JuliaMath/Polynomials.jl Polynomials.jl@noreply.github.com Cc: B-LIE bernt.lie@hotmail.com; Author author@noreply.github.com Subject: Re: [JuliaMath/Polynomials.jl] Limits of Polynomials.PolyCompat? (#247)

Wow, that is subtle, and sadly consistent with the other report that was related. There a new gaming machine was the only machine that caused issues. I'll see if I can recreate locally somehow so I can slowly peel back the plot commands. There must be something triggering this.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://eur06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FJuliaMath%2FPolynomials.jl%2Fissues%2F247%23issuecomment-667068677&data=02%7C01%7C%7C331ae5874c5c44fb006708d83542bc05%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317908014707317&sdata=uQIAvHtY12%2BiQeJM3zmEDMBJhgpd%2FQ2UYIDogiOEgug%3D&reserved=0, or unsubscribehttps://eur06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAM5VM76BWZ5R3OQXONLMV6DR6KRNBANCNFSM4PLYPZ4Q&data=02%7C01%7C%7C331ae5874c5c44fb006708d83542bc05%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637317908014717308&sdata=73egzYHUQQaYKAt3ualrHw3Y1%2BkbU9I4ar1c%2FjQngHM%3D&reserved=0.

jverzani commented 4 years ago

In the other reported issue (#244) there was a report up to pyplot ( https://github.com/JuliaPy/PyPlot.jl/issues/477) which sadly was inconclusive. That user had this comment:

So perhaps it is CPU related, perhaps not. The remark in the response to the issue that it likely sits at an intersection of python's numpy and LinearAlgebra's different use of LAPACK is likely where this bug lies. My thought that LaTeX was somehow involved is dispelled by the opener of the issue, who shows only that a call to xlim can trigger this.

On Fri, Jul 31, 2020 at 7:23 AM B-LIE notifications@github.com wrote:

-- John Verzani Department of Mathematics College of Staten Island, CUNY verzani@math.csi.cuny.edu

JuliaMath / Polynomials.jl

Limits of `Polynomials.PolyCompat`? #247

I’m trying to install one-by-one additional packages on my old desktop that were lacking in the new one. I’ll report when I have tested the remaining 4 packages on the old desktop.