Pattern with index and set restriction matches to number

[benruijl]

The following id statement matches even though it shouldn't:

#-
I mu1,mu2,mu3;
CF f;
Set indices:mu1,mu2,mu3;

L F = f(132); * only happens for certain numbers!
id f(mu1?indices) = 1; * matches...

Print +s;
.end

[tueda]

As you may imagine

On code;
I mu1,mu2,mu3;
.end

prints

 Indices
   iarg_(129) mu1(130) mu2(131) mu3(132)

[tueda]

A (silly) workaround is putting

#:constindex 10000

[tueda]

To me, the problem (also in #149) is resolved if we forbid a small number function argument that is not in the range of fixed indices matching with any indices at function.c:1242:

diff --git a/sources/function.c b/sources/function.c
index 7ec75dd..87fc8f0 100644
--- a/sources/function.c
+++ b/sources/function.c
@@ -1239,7 +1239,8 @@ IndAll:                           i = m[1] - WILDOFFSET;
                        else if ( *m == -ARGWILD ) goto ArgAll;
                        else if ( *m == -INDEX && m[1] >= AM.OffsetIndex+WILDOFFSET
                        && m[1] < AM.OffsetIndex+(WILDOFFSET<<1) ) {
-                               if ( *t == -VECTOR || *t == -SNUMBER ) goto IndAll;
+                               if ( *t == -VECTOR ) goto IndAll;
+                               if ( *t == -SNUMBER && t[1] >= 0 && t[1] < AM.OffsetIndex ) goto IndAll;
                                if ( *t == -MINVECTOR ) {
                                        i = m[1] - WILDOFFSET;
                                        AN.argaddress = AT.MinVecArg;

Am I right? I'm still not sure about the fixed indices...

[vermaseren]

Hi Takahiro,

That sounds about right. It is probably just a simple if somewhere in the code. I am a bit too busy with the colors at the moment. There is a nontrivial bug, either in color.h or in FORM. This is far more serious.

Jos

On 28 nov. 2016, at 13:49, Takahiro Ueda notifications@github.com wrote:

To me, the problem (also in #149 https://github.com/vermaseren/form/issues/149) is resolved if we forbid a small number function argument that is not in the range of fixed indices matching with any indices at function.c:1242 https://github.com/vermaseren/form/blob/6fe5632f513100fcabdc4c7fb002609b1bf8efea/sources/function.c#L1242:

diff --git a/sources/function.c b/sources/function.c index 7ec75dd..87fc8f0 100644 --- a/sources/function.c +++ b/sources/function.c @@ -1239,7 +1239,8 @@ IndAll: i = m[1] - WILDOFFSET; else if ( m == -ARGWILD ) goto ArgAll; else if ( m == -INDEX && m[1] >= AM.OffsetIndex+WILDOFFSET && m[1] < AM.OffsetIndex+(WILDOFFSET<<1) ) {

• if ( t == -VECTOR || t == -SNUMBER ) goto IndAll;
• if ( *t == -VECTOR ) goto IndAll;
• if ( t == -SNUMBER && t[1] >= 0 && t[1] < AM.OffsetIndex ) goto IndAll; if ( t == -MINVECTOR ) { i = m[1] - WILDOFFSET; AN.argaddress = AT.MinVecArg; Am I right? I'm still not sure about the fixed indices...

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/vermaseren/form/issues/153#issuecomment-263263281, or mute the thread https://github.com/notifications/unsubscribe-auth/AFLxEpAEEjq_07JGTs64VYynEs4X4C2Sks5rCs3tgaJpZM4K84aJ.

[tueda]

Hi Jos,

OK. I will commit this patch, which fixes at least these cases. We will see if there are more related bugs.