Msolve #3366

antonleykin · 2024-07-17T22:32:26Z

This introduces Msolve package, which is an interface to msolve developed mainly by @Martin-Helmer (with parsing routines in the kernel by @mikestillman).

M2/Macaulay2/packages/Msolve.m2

M2/Macaulay2/packages/Msolve/EXA/Msolve-testing.m2

mikestillman · 2024-07-18T12:33:19Z

@d-torrance Is it possible to have the autotools build automatically install msolve? @mahrud I believe we can do this (or it is already done) on the cmake build? (Here, I mean the msolve executable).

d-torrance · 2024-07-18T12:36:54Z

@d-torrance Is it possible to have the autotools build automatically install msolve?

Certainly! Would you like me to write the code to do this?

mikestillman · 2024-07-18T12:37:31Z

@d-torrance That would be great if you could do so! Thanks!

d-torrance · 2024-07-18T13:34:58Z

@mikestillman - See antonleykin#26

mahrud · 2024-07-18T14:29:53Z

@d-torrance you could also just push here since it says "Maintainers are allowed to edit this pull request." Then we have one fewer merge commit ...

d-torrance · 2024-07-18T14:33:19Z

Ah, indeed! Done.

mahrud · 2024-07-18T14:33:35Z

I believe we can do this (or it is already done) on the cmake build?

Yes, it already happens.

d-torrance · 2024-07-19T12:20:01Z

I just pushed a couple more commits, adding the package to =distributed-packages (which should fix the broken link error), and installing the Ubuntu/brew msolve packages for the GitHub builds.

mikestillman · 2024-07-19T12:36:00Z

@d-torrance @mahrud What versions are msolve at in ubuntu/homebrew/M2? Mohab reported that a bug I noticed (in version v0.6.5) appears to be fixed in v0.6.6.

d-torrance · 2024-07-19T12:42:25Z

For Ubuntu, it depends on the release. Ubuntu 24.04 has 0.6.5, and 24.10 will ship with 0.6.6. See https://launchpad.net/ubuntu/+source/msolve. I can always update the version in the PPA (which covers LTS releases all the way back to 18.04) -- right now it's still at 0.4.9.

I'd forgotten that there was an 0.6.6 release -- it's not listed on the main msolve webpage, just their GitHub page. I should update the autotools commit to build that instead.

d-torrance · 2024-07-19T20:24:52Z

I think the Ubuntu build errors might be coming from the old version of msolve (0.4.9) in the PPA. I'm working on packaging the newest version (0.6.7, released earlier today) for Debian right now. I'll backport it for the PPA after that, and hopefully that will fix the builds.

Martin-Helmer · 2024-07-19T20:36:01Z

I think the Ubuntu build errors might be coming from the old version of msolve (0.4.9) in the PPA. I'm working on packaging the newest version (0.6.7, released earlier today) for Debian right now. I'll backport it for the PPA after that, and hopefully that will fix the builds.

Yes, I would expect the tests in the Msolve M2 package to fail with any msolve version before 0.6.4

d-torrance · 2024-07-20T01:59:35Z

msolve 0.6.7 is now live on the PPA, so I've restarted the Ubuntu builds.

M2/Macaulay2/packages/Msolve.m2

mahrud · 2024-07-21T00:31:48Z

M2/Macaulay2/packages/Msolve.m2

+    if opts#"output type"=="rationalInterval" then return sols;
+    if opts#"output type"=="floatInterval" then return (1.0*sols);
+    if opts#"output type"=="float" then return (for s in sols list(for s1 in s list sum(s1)/2.0));


I'm guessing "float"=RR and "floatInterval"=RRi, but what is "rationalInterval"? Also, isn't it easier for the user to do 1.0*sols if they want, and use midpoint for the "float" option?

Here are a few alternatives that I think would be more idiomatic here:

split this function into msolveRealSolution, msolveRealIntervalSolution, and msolveRationalSolutionInterval (if this exists in M2)

have this function always return sols, i.e. "rationalInterval" option, but have helper functions like midpoint for turning that into a float interval or float, which could also be used in other places idiomatically

add a new input of type Ring to this function, which could be QQ or RR or RRi and cast to that ring.

I don't have a strong feeling here, but mostly the interval return might be a bit hard for people to read. But it makes sense for the default as it is still the most accurate answer.

I don't like suggestion 3 so much as expect feeding anything other than QQ or a finite field into msolve is going to create infinite problems.

I agree that the default should be the most accurate, i.e. sols. Why is it not returning RRi interval objects?

For option 3, it should definitely throw an error if the ring is anything other than the handful of fields we can handle.

I am not returning RRi objects mostly since I am not familiar with these.

The default return is what msolve would return, unedited. So it is intervals of rationals.

I have left this as is in my most recent update, but if others want to change it that is also fine with me.

M2/Macaulay2/packages/Macaulay2Doc/overview3.m2

M2/Macaulay2/packages/Msolve.m2

mahrud · 2024-07-21T00:48:45Z

M2/Macaulay2/packages/Msolve.m2

+    );
+msolveLeadMonomials=method(TypicalValue=>Matrix, Options=>{"level of verbosity"=>0, "number of threads"=>maxAllowableThreads});
+msolveLeadMonomials(Ideal):=opt->(I)->(
+    if not inputOkay(I) then return 0;


Same, throwing an error. We can add hooks for leadMonomial and add this as a hook, which can be specified using the Strategy.

Also, should this be called msolveLeadTerm instead, or am I missing something?

msolve always returns monic monomials with the GB option which computes the lead term. So this seems like a reasonable name?

Aren't the lead terms of a Grobner basis over a field are always monic? e.g. in M2 leadTerm gb I always returns monic monomials also, so I think it's important to align the methods with M2's native ones.

leadTerm groebnerBasis I
and
leadTerm I
both return a matrix with the coefficients included (and are not necessarily monic).

And for me leadTerm gb I does not nessarily return monic monomials (of course you can just throw away the coefficients, but that is not what it does, and leadTerm seems more correct when you keep the coefficients... the ideal is of course the same either way).

Anyway, to me the msolve GB is closer aligned to the "groebnerBasis" command rather than the "gb" command, which was why I opted to do it this way, i.e. if it was msolveLeadTerm I would want its output to match leadTerm groebnerBasis I, but it won't.

Also msolve only has grevlex and block grevlex (as an elimination order) defined, so in some sense I am not sure we want to align it too closely to the M2 commands since it has more limited capabilities in some sense.

Ah interesting, it seems like this is actually field dependent:

i21 : R = ZZ[x,y]; leadTerm gb ideal (2*x+3*y) o22 = | 2x | 1 1 o22 : Matrix R <-- R i23 : R = ZZ/5[x,y]; leadTerm gb ideal (2*x+3*y) o24 = | x | 1 1 o24 : Matrix R <-- R

@mikestillman is this a bug?

M2/Macaulay2/packages/Msolve.m2

mahrud · 2024-07-21T00:59:38Z

M2/Macaulay2/packages/Msolve.m2

+msolvePresentAndModern := msolvePresent -- and match("^[0-9.]+$",msolveVersion) and msolveVersion >= msolveVersionNeeded
+
+newPackage(
+	"Msolve",


Should this be capitalized as MSolve instead? e.g. see the Sage interface.

the msolve team always seems to call it msolve or MSOLVE.... so... I think there is no canonical choice.

Other alternative is msolveInterface (following gfanInterface). Eventually, it might be a good idea to have a subdirectory packages/interfaces which contains all interface packages named exactly like the software, e.g. gfan.m2 msolve.m2 etc.

Thoughts @mikestillman?

mahrud

Could you shorten these headlines?

[1/1] Installing package Msolve
 -- warning: document: documentation headlines must be less than 100 characters long:
 --   "Macaulay2 interface for msolve; computes real solutions, Groebner basis (in GRevLex), saturations, and elimination ideals."
 -- warning: document: documentation headlines must be less than 100 characters long:
 --   "Uses symbolic methods to compute a certified solution interval for each real solution to a zero dimensional polynomial system."
 -- warning: document: documentation headlines must be less than 100 characters long:
 --   "Uses symbolic methods to compute the rational univariate representation of a zero dimensional polynomial system."
 -- warning: document: documentation headlines must be less than 100 characters long:
 --   "Computes a Groebner basis for the saturation of an ideal by a single polynomial in GrevLex order; only works over a finite field."

From Package Writing Style Guide:

The "headline" of a documentation node should be a single brief phrase (not a complete sentence) and thus can be vague. It should not refer to variables by name, because it can appear in a menu, without the accompanying body of the documentation node.

stdout may still not print until the end if it isn't flushed. Also, use printerr for "running" message and display exit status on error.

d-torrance · 2024-08-05T18:55:01Z

I've fixed the runProgram issues I mentioned above and also rebased this branch on top of development, resolving the conflicts.

mahrud · 2024-08-05T19:21:13Z

M2/Macaulay2/m2/programs.m2

+    returnValue := (
+	if opts.Verbose
+	then run ("(((((" | cmd | "; echo $? >&4) | tee " | outFile |
+	    ") 3>&1 1>&2 2>&3 | tee " | errFile |
+	    " >&5) 4>&1) | (read r; exit $r)) 5>&1")
+	else run (cmd | " > " | outFile | " 2> " | errFile));


This works great, thank you!
For future though, I wonder if it would be best to add a c++ routine for this.

mahrud · 2024-08-07T05:59:46Z

GitHub hid this message after the rebase, so quoting final thoughts from above:

perhaps renaming to msolveInterface.m2, which matches their own capitalization of msolve as well as the package gfanInterface.m2 is more appropriate.
I still think the optional input of msolveRealSolutions is strange, and instead it should return an RRi interval, which the user can take midpoint of, for instance.

antonleykin · 2024-08-07T13:49:19Z

GitHub hid this message after the rebase, so quoting final thoughts from above:

perhaps renaming to msolveInterface.m2, which matches their own capitalization of msolve as well as the package gfanInterface.m2 is more appropriate.

I still think the optional input of msolveRealSolutions is strange, and instead it should return an RRi interval, which the user can take midpoint of, for instance.

The majority of distributed "Interfaces" follow the pattern that @Martin-Helmer followed. I see no problem.

Capitalization is done the same way for nauty, bertini, topcom, ...

mahrud · 2024-08-07T14:07:16Z

Okay. I'll merge this then, and maybe we can come back to msolveRealSolutions later.

pzinn · 2024-08-08T06:50:11Z

M2/Macaulay2/packages/Msolve.m2

+if msolveProgram === null then (
+    if not (options currentPackage).OptionalComponentsPresent
+    then ( printerr "warning: msolve cannot be found; ending"; end )
+    else ( printerr "warning: msolve found but its version is older than v" | msolveMinimumVersion;


This line won't run, you need parentheses around the argument. (a bit late I know)

d-torrance reviewed Jul 18, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

d-torrance reviewed Jul 18, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve/EXA/Msolve-testing.m2 Outdated Show resolved Hide resolved

d-torrance force-pushed the Msolve branch 2 times, most recently from 8fc56b2 to d8046ad Compare July 19, 2024 12:17

d-torrance reviewed Jul 20, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

d-torrance reviewed Jul 20, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

d-torrance mentioned this pull request Jul 20, 2024

Document "Keywords" symbol #3368

Closed

mahrud reviewed Jul 21, 2024

View reviewed changes

M2/Macaulay2/packages/Macaulay2Doc/overview3.m2 Show resolved Hide resolved

mahrud reviewed Jul 21, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

mahrud mentioned this pull request Jul 21, 2024

Move OptionalComponentsPresent from newPackage to beginDocumentation #3369

Open

mahrud reviewed Jul 21, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

mahrud reviewed Jul 21, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

mahrud reviewed Jul 21, 2024

View reviewed changes

M2/Macaulay2/packages/Msolve.m2 Outdated Show resolved Hide resolved

mahrud reviewed Jul 21, 2024

View reviewed changes

mahrud requested changes Jul 21, 2024

View reviewed changes

mahrud and others added 19 commits August 5, 2024 14:54

added characteristic check for msolve bug

985082b

added a test for msolveSaturate

c78d0d4

simplified msolveSaturate

ee6ba11

further cosmetic documentation changes

c838eda

unified ring sanitization for msolve

f6c08fe

added helper routine readMsolveList

903c1fe

fixed an issue in monoidIndex

2af9f95

added kludge for weird monomial order bug

42d2b50

added more ring tower tests for msolve

d931c5b

fixed formatting of readMsolve unit-test

e7cf5ed

removed the caveat for msolveSaturate

8dc4a06

bumped minimum msolve version to 0.7.0

cd87eaa

bumped msolve to version 0.7.1 (cmake)

8630b34

removed OptionalComponentsPresent from Msolve

5c03e20

fixed a bug and added benchmarking data

c9eed72

fixed msolve url in cmake

3b2f516

added two more msolve benchmarks

af0990d

combined msolve tests and added assertions

7cbe2b5

Print verbose runProgram output as it happens

f0a430d

stdout may still not print until the end if it isn't flushed. Also, use printerr for "running" message and display exit status on error.

d-torrance force-pushed the Msolve branch from a35f2a0 to f0a430d Compare August 5, 2024 18:54

mahrud reviewed Aug 5, 2024

View reviewed changes

mahrud merged commit bf9ac3e into Macaulay2:development Aug 7, 2024
5 checks passed

mahrud mentioned this pull request Aug 7, 2024

MSolve for multivariate polynomial system solving #2132

Closed

antonleykin mentioned this pull request Aug 7, 2024

msolveRealSolutions #3392

Merged

pzinn reviewed Aug 8, 2024

View reviewed changes

d-torrance mentioned this pull request Oct 19, 2024

Bump TerraciniLoci package to version 0.2 #3527

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Msolve #3366

Msolve #3366

antonleykin commented Jul 17, 2024

mikestillman commented Jul 18, 2024

d-torrance commented Jul 18, 2024

mikestillman commented Jul 18, 2024

d-torrance commented Jul 18, 2024

mahrud commented Jul 18, 2024

d-torrance commented Jul 18, 2024

mahrud commented Jul 18, 2024

d-torrance commented Jul 19, 2024

mikestillman commented Jul 19, 2024

d-torrance commented Jul 19, 2024

d-torrance commented Jul 19, 2024 •

edited

Loading

Martin-Helmer commented Jul 19, 2024

d-torrance commented Jul 20, 2024

mahrud Jul 21, 2024

Martin-Helmer Jul 23, 2024

mahrud Jul 23, 2024

Martin-Helmer Jul 25, 2024 •

edited

Loading

mahrud Jul 21, 2024

mahrud Jul 22, 2024

Martin-Helmer Jul 23, 2024

mahrud Jul 23, 2024

Martin-Helmer Jul 23, 2024

mahrud Jul 25, 2024

mahrud Jul 21, 2024

Martin-Helmer Jul 23, 2024

mahrud Jul 23, 2024

mahrud left a comment

d-torrance commented Aug 5, 2024

mahrud Aug 5, 2024

mahrud commented Aug 7, 2024

antonleykin commented Aug 7, 2024 •

edited

Loading

mahrud commented Aug 7, 2024

pzinn Aug 8, 2024

Msolve #3366

Msolve #3366

Conversation

antonleykin commented Jul 17, 2024

mikestillman commented Jul 18, 2024

d-torrance commented Jul 18, 2024

mikestillman commented Jul 18, 2024

d-torrance commented Jul 18, 2024

mahrud commented Jul 18, 2024

d-torrance commented Jul 18, 2024

mahrud commented Jul 18, 2024

d-torrance commented Jul 19, 2024

mikestillman commented Jul 19, 2024

d-torrance commented Jul 19, 2024

d-torrance commented Jul 19, 2024 • edited Loading

Martin-Helmer commented Jul 19, 2024

d-torrance commented Jul 20, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Martin-Helmer Jul 25, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mahrud left a comment

Choose a reason for hiding this comment

d-torrance commented Aug 5, 2024

Choose a reason for hiding this comment

mahrud commented Aug 7, 2024

antonleykin commented Aug 7, 2024 • edited Loading

mahrud commented Aug 7, 2024

Choose a reason for hiding this comment

d-torrance commented Jul 19, 2024 •

edited

Loading

Martin-Helmer Jul 25, 2024 •

edited

Loading

antonleykin commented Aug 7, 2024 •

edited

Loading