g_scalar.c

/* Copyright (c) 1997-1999 Miller Puckette.
* For information on usage and redistribution, and for a DISCLAIMER OF ALL
* WARRANTIES, see the file, "LICENSE.txt," in this distribution.  */

/* This file defines the "scalar" object, which is not a text object, just a
"gobj".  Scalars have templates which describe their structures, which
can contain numbers, sublists, and arrays.

*/

#include <stdlib.h>
#include <string.h>
#include <stdio.h>      /* for read/write to files */
#include "m_pd.h"
#include "g_canvas.h"

t_class *scalar_class;

void word_init(t_word *wp, t_template *template, t_gpointer *gp)
{
    int i, nitems = template->t_n;
    t_dataslot *datatypes = template->t_vec;
    for (i = 0; i < nitems; i++, datatypes++, wp++)
    {
        int type = datatypes->ds_type;
        if (type == DT_FLOAT)
            wp->w_float = 0; 
        else if (type == DT_SYMBOL)
            wp->w_symbol = &s_symbol;
        else if (type == DT_ARRAY)
        {
            wp->w_array = array_new(datatypes->ds_arraytemplate, gp);
        }
        else if (type == DT_LIST)
        {
                /* LATER test this and get it to work */
            wp->w_list = canvas_new(0, 0, 0, 0);
        }
    }
}

void word_restore(t_word *wp, t_template *template,
    int argc, t_atom *argv)
{
    int i, nitems = template->t_n;
    t_dataslot *datatypes = template->t_vec;
    for (i = 0; i < nitems; i++, datatypes++, wp++)
    {
        int type = datatypes->ds_type;
        if (type == DT_FLOAT)
        {
            t_float f;
            if (argc)
            {
                f =  atom_getfloat(argv);
                argv++, argc--;
            }
            else f = 0;
            wp->w_float = f; 
        }
        else if (type == DT_SYMBOL)
        {
            t_symbol *s;
            if (argc)
            {
                s =  atom_getsymbol(argv);
                argv++, argc--;
            }
            else s = &s_;
            wp->w_symbol = s;
        }
    }
    if (argc)
        post("warning: word_restore: extra arguments");
}

void word_free(t_word *wp, t_template *template)
{
    int i;
    t_dataslot *dt;
    for (dt = template->t_vec, i = 0; i < template->t_n; i++, dt++)
    {
        if (dt->ds_type == DT_ARRAY)
            array_free(wp[i].w_array);
        else if (dt->ds_type == DT_LIST)
            canvas_free(wp[i].w_list);
    }
}

/* some of this code is used in a function in g_canvas.c...
   need to modularize it */
static t_object *template_getstruct(t_template *template)
{
    if (template)
    {
        t_gobj *y;
        t_canvas *c;
        if (c = template_findcanvas(template))
        {
            t_symbol *s1 = gensym("struct");
            for (y = c->gl_list; y; y = y->g_next)
            {
                t_object *ob = pd_checkobject(&y->g_pd);
                t_atom *argv;
                if (!ob || ob->te_type != T_OBJECT ||
                    binbuf_getnatom(ob->te_binbuf) < 2)
                    continue;
                argv = binbuf_getvec(ob->te_binbuf);
                if (argv[0].a_w.w_symbol != s1)
                    continue;
                if (canvas_makebindsym(argv[1].a_w.w_symbol) == template->t_sym)
                    return (ob);
            }
        }
    }
    return (0);
}

int template_hasxy(t_template *template)
{
    t_symbol *zz;
    int xonset, yonset, xtype, ytype, gotx, goty;
    if (!template)
    {
        error("struct: couldn't find template %s", template->t_sym->s_name);
        return 0;
    }
    gotx = template_find_field(template, gensym("x"), &xonset, &xtype, &zz);
    goty = template_find_field(template, gensym("y"), &yonset, &ytype, &zz);
    if ((gotx && (xtype == DT_FLOAT)) &&
        (goty && (ytype == DT_FLOAT)) &&
        (xonset == 0) && (yonset == sizeof(t_word)))
    {
        return 1;
    }
    else
        return 0;
}

int template_cancreate(t_template *template)
{
    int i, type, nitems = template->t_n;
    t_dataslot *datatypes = template->t_vec;
    t_template *elemtemplate;
    for (i = 0; i < nitems; i++, datatypes++)
        if (datatypes->ds_type == DT_ARRAY &&
            (!(elemtemplate = template_findbyname(datatypes->ds_arraytemplate))
                || !template_cancreate(elemtemplate)))
    {
        t_object *ob = template_getstruct(template);
        pd_error(ob, "%s: no such template",
            datatypes->ds_arraytemplate->s_name);
        return (0);
    }
    return (1);
}

    /* make a new scalar and add to the glist.  We create a "gp" here which
    will be used for array items to point back here.  This gp doesn't do
    reference counting or "validation" updates though; the parent won't go away
    without the contained arrays going away too.  The "gp" is copied out
    by value in the word_init() routine so we can throw our copy away. */

t_scalar *scalar_new(t_glist *owner, t_symbol *templatesym)
{
    t_scalar *x;
    t_template *template;
    t_gpointer gp;
    gpointer_init(&gp);
    template = template_findbyname(templatesym);
    if (!template)
    {
        error("scalar: couldn't find template %s", templatesym->s_name);
        return (0);
    }
    if (!template_cancreate(template))
        return (0);
    x = (t_scalar *)getbytes(sizeof(t_scalar) +
        (template->t_n - 1) * sizeof(*x->sc_vec));
    x->sc_gobj.g_pd = scalar_class;
    x->sc_template = templatesym;
    gpointer_setglist(&gp, owner, x);
    word_init(x->sc_vec, template, &gp);
    char buf[50];
    sprintf(buf, ".x%lx", (long unsigned int)x);
    pd_bind(&x->sc_gobj.g_pd, gensym(buf));
    return (x);
}

    /* Pd method to create a new scalar, add it to a glist, and initialize
    it from the message arguments. */

int glist_readscalar(t_glist *x, int natoms, t_atom *vec,
    int *p_nextmsg, int selectit);

void glist_scalar(t_glist *glist,
    t_symbol *classname, t_int argc, t_atom *argv)
{
    t_symbol *templatesym =
        canvas_makebindsym(atom_getsymbolarg(0, argc, argv));
    t_binbuf *b;
    int natoms, nextmsg = 0;
    t_atom *vec;
    if (!template_findbyname(templatesym))
    {
        pd_error(glist, "%s: no such template",
            atom_getsymbolarg(0, argc, argv)->s_name);
        return;
    }

    b = binbuf_new();
    binbuf_restore(b, argc, argv);
    natoms = binbuf_getnatom(b);
    vec = binbuf_getvec(b);
    
    glist_readscalar(glist, natoms, vec, &nextmsg, 0);
    binbuf_free(b);
}

    /* search template fields recursively to see if the template
       depends on elemtemplate */
int template_has_elemtemplate(t_template *t, t_template *elemtemplate)
{
    int returnval = 0;
    if (t && elemtemplate)
    {
        int i;
	t_dataslot *d = t->t_vec;
        for (i = 0; i < t->t_n; i++, d++)
        {
            if (d->ds_type == DT_ARRAY)
            {
                if (d->ds_arraytemplate == elemtemplate->t_sym)
                {
                    returnval = 1;
                    break;
                }
                else
                {
                    returnval = template_has_elemtemplate(
                        template_findbyname(d->ds_arraytemplate),
                        elemtemplate);
                }
            }
        }
    }
    return (returnval);
}

/* -------------------- widget behavior for scalar ------------ */
void scalar_getbasexy(t_scalar *x, t_float *basex, t_float *basey)
{
    t_template *template = template_findbyname(x->sc_template);
    *basex = template_getfloat(template, gensym("x"), x->sc_vec, 0);
    *basey = template_getfloat(template, gensym("y"), x->sc_vec, 0);
}

extern int array_joc;

extern void template_notifyforscalar(t_template *template, t_glist *owner,
    t_scalar *sc, t_symbol *s, int argc, t_atom *argv);

t_canvas *sc_mouseover_canvas;
static void scalar_mouseover(t_scalar *x, t_floatarg state)
{
    t_atom at[1];
    t_template *template = template_findbyname(x->sc_template);
    if (state)
        template_notifyforscalar(template, sc_mouseover_canvas,
            x, gensym("leave"), 1, at);
    else
        template_notifyforscalar(template, sc_mouseover_canvas,
            x, gensym("enter"), 1, at);
}

static void scalar_getrect(t_gobj *z, t_glist *owner,
    int *xp1, int *yp1, int *xp2, int *yp2)
{
    //fprintf(stderr,"scalar_getrect %d\n", array_joc);
    t_scalar *x = (t_scalar *)z;

    t_template *template = template_findbyname(x->sc_template);
    t_canvas *templatecanvas = template_findcanvas(template);
    int x1 = 0x7fffffff, x2 = -0x7fffffff, y1 = 0x7fffffff, y2 = -0x7fffffff;
    t_gobj *y;
    t_float basex, basey;

    // EXPERIMENTAL: we assume that entire canvas is within
    // the rectangle--this is for arrays
    // with "jump on click" enabled TODO: test for other regressions
    // (there shouuld not be any
    // provided the global variable array_joc is properly maintained)
    if (glist_istoplevel(owner) && array_joc)
    {
        x1 = -0x7fffffff, y1 = -0x7fffffff, x2 = 0x7fffffff, y2 = 0x7fffffff;
    }

    else
    {
        scalar_getbasexy(x, &basex, &basey);
            /* if someone deleted the template canvas, we're just a point */
        if (!templatecanvas)
        {
            //fprintf(stderr,"...point\n");
            x1 = x2 = glist_xtopixels(owner, basex);
            y1 = y2 = glist_ytopixels(owner, basey);
        }
        else
        {
            /* todo: bad flow with internal return here. make it cleaner */
            if (x->sc_bboxcache)
            {
                *xp1 = x->sc_x1;
                *yp1 = x->sc_y1;
                *xp2 = x->sc_x2;
                *yp2 = x->sc_y2;
                return;
            }
            x1 = y1 = 0x7fffffff;
            x2 = y2 = -0x7fffffff;
            for (y = templatecanvas->gl_list; y; y = y->g_next)
            {
                t_parentwidgetbehavior *wb = pd_getparentwidget(&y->g_pd);
                int nx1, ny1, nx2, ny2;
                if (!wb) continue;
                (*wb->w_parentgetrectfn)(y, owner,
                    x->sc_vec, template, basex, basey,
                    &nx1, &ny1, &nx2, &ny2);
                if (nx1 < x1) x1 = nx1;
                if (ny1 < y1) y1 = ny1;
                if (nx2 > x2) x2 = nx2;
                if (ny2 > y2) y2 = ny2;
                //fprintf(stderr,"====scalar_getrect x1 %d y1 %d x2 %d y2 %d\n",
                //    x1, y1, x2, y2);
            }
            if (x2 < x1 || y2 < y1)
                x1 = y1 = x2 = y2 = 0;
        }
    }
    //fprintf(stderr,"FINAL scalar_getrect x1 %d y1 %d x2 %d y2 %d\n",
    //    x1, y1, x2, y2);
    *xp1 = x->sc_x1 = x1;
    *yp1 = x->sc_y1 = y1;
    *xp2 = x->sc_x2 = x2;
    *yp2 = x->sc_y2 = y2; 
    x->sc_bboxcache = 1;
}

void scalar_drawselectrect(t_scalar *x, t_glist *glist, int state)
{
    //fprintf(stderr,"scalar_drawselecterect%d\n", state);
    if (state)
    {
        int x1, y1, x2, y2;
       
        scalar_getrect(&x->sc_gobj, glist, &x1, &y1, &x2, &y2);
        x1--; x2++; y1--; y2++;
                /* we're not giving the rectangle the "select" tag
                   because we have to manually displace the scalar
                   in scalar_displace_withtag. The reason for that
                   is the "displace" message may trigger a redraw
                   of the bbox at the new position, and Pd-l2ork's
                   general "move selected" subcommand will end up
                   offsetting such a rect by dx dy.
                */
        if (glist_istoplevel(glist))
            sys_vgui(".x%lx.c create prect %d %d %d %d "
                     "-strokewidth 1 -stroke $pd_colors(selection) "
                     "-tags {select%lx}\n",
                    glist_getcanvas(glist), x1, y1, x2, y2,
                    x);
    }
    else
    {
        if (glist_istoplevel(glist))
            sys_vgui(".x%lx.c delete select%lx\n", glist_getcanvas(glist), x);
    }
}

/* This is a workaround.  Since scalars are contained within a tkpath
   group, and since tkpath groups don't have coords, we have to fudge
   things with regard to Pd-l2ork's normal *_displace_withtag functionality.
   (Additionally, we can't just fall back to the old displace method
   because it too assumes the canvas item has an xy coord.)

   We draw the selection rect but don't add the scalar to the
   "selected" tag.  This means when Pd-l2ork issues the canvas "move"
   command, our scalar doesn't go anywhere.  Instead we get the callback
   to scalar_displace_withtag and do another workaround to get the
   new position and feed it to the scalar's matrix.

   This creates a problem with gop canvases, and yet _another_ partial
   workaround which I apologize for inside t_scalar def in m_pd.h.
*/
void scalar_select(t_gobj *z, t_glist *owner, int state)
{
    //fprintf(stderr,"scalar_select %d\n", state);
    t_scalar *x = (t_scalar *)z;
    t_template *tmpl;
    t_symbol *templatesym = x->sc_template;
    t_atom at;
    //t_canvas *templatecanvas = NULL;
    t_gpointer gp;
    gpointer_init(&gp);
    gpointer_setglist(&gp, owner, x);
    SETPOINTER(&at, &gp);
    if (tmpl = template_findbyname(templatesym))
    {
        template_notify(tmpl, (state ? gensym("select") : gensym("deselect")),
            1, &at);
        //templatecanvas = template_findcanvas(tmpl);
    }
    gpointer_unset(&gp);
    if (state)
    {
        x->sc_selected = owner;
        sys_vgui(".x%lx.c addtag selected withtag blankscalar%lx\n",
            glist_getcanvas(owner), x);
        /* how do we navigate through a t_word list?
        if (x->sc_vec)
        {
            t_word *v = x->sc_vec;
            while(v)
            {
                sys_vgui(".x%lx.c "
                         "addtag selected withtag .x%lx.x%lx.template%lx\n",
                    glist_getcanvas(owner), glist_getcanvas(owner), owner, v);
            }
        }*/
        /*if (templatecanvas) {
            // get the universal tag for all nested objects
            t_canvas *tag = owner;
            while (tag->gl_owner)
            {
                tag = tag->gl_owner;
            }
            sys_vgui(".x%lx.c addtag selected withtag %lx\n",
                glist_getcanvas(owner), (t_int)tag);
        }*/
    }
    else
    {
        x->sc_selected = 0;
        sys_vgui(".x%lx.c dtag blankscalar%lx selected\n",
            glist_getcanvas(owner), x);
        sys_vgui(".x%lx.c dtag .x%lx.x%lx.template%lx selected\n",
            glist_getcanvas(owner), glist_getcanvas(owner),
            owner, x->sc_vec);
        /* how do we navigate through a t_word list?
        if (x->sc_vec)
        {
            t_word *v = x->sc_vec;
            while (v)
            {
                sys_vgui(".x%lx.c dtag .x%lx.x%lx.template%lx selected\n",
                    glist_getcanvas(owner), glist_getcanvas(owner),
                    owner, x->sc_vec);
            }
        }*/
        /*if (templatecanvas) {
            // get the universal tag for all nested objects
            t_canvas *tag = owner;
            while (tag->gl_owner)
            {
                tag = tag->gl_owner;
            }
            sys_vgui(".x%lx.c dtag %lx selected\n",
                glist_getcanvas(owner), (t_int)tag);
        }*/
    }
    //sys_vgui("pdtk_select_all_gop_widgets .x%lx %lx %d\n",
    //    glist_getcanvas(owner), owner, state);
    scalar_drawselectrect(x, owner, state);
}

static void scalar_displace(t_gobj *z, t_glist *glist, int dx, int dy)
{
    //fprintf(stderr,"scalar_displace\n");
    t_scalar *x = (t_scalar *)z;
    t_symbol *templatesym = x->sc_template;
    t_template *template = template_findbyname(templatesym);
    t_symbol *zz;
    t_atom at[3];
    t_gpointer gp;
    int xonset, yonset, xtype, ytype, gotx, goty;
    t_float basex = 0, basey = 0;
    if (!template)
    {
        error("scalar: couldn't find template %s", templatesym->s_name);
        return;
    }
    gotx = template_find_field(template, gensym("x"), &xonset, &xtype, &zz);
    if ((gotx && (xtype != DT_FLOAT)) || x->sc_selected != glist)
        gotx = 0;
    goty = template_find_field(template, gensym("y"), &yonset, &ytype, &zz);
    if ((goty && (ytype != DT_FLOAT)) || x->sc_selected != glist)
        goty = 0;
    if (gotx)
    {
        *(t_float *)(((char *)(x->sc_vec)) + xonset) +=
            dx * (glist_pixelstox(glist, 1) - glist_pixelstox(glist, 0));
        x->sc_x1 += dx;
        x->sc_x2 += dx;
    }
    if (goty)
    {
        *(t_float *)(((char *)(x->sc_vec)) + yonset) +=
            dy * (glist_pixelstoy(glist, 1) - glist_pixelstoy(glist, 0));
        x->sc_y1 += dy;
        x->sc_y2 += dy;
    }
    gpointer_init(&gp);
    gpointer_setglist(&gp, glist, x);
    SETPOINTER(&at[0], &gp);
    SETFLOAT(&at[1], (t_float)dx);
    SETFLOAT(&at[2], (t_float)dy);
    template_notify(template, gensym("displace"), 2, at);
    scalar_redraw(x, glist);
}

/* Very complicated at the moment. If a scalar is in a gop canvas, then
   we don't need to update its x/y fields (if it even has them) when displacing
   it.  Otherwise we do.  The global selected_owner variable is used to store
   the "owner" canvas-- if it matches the glist parameter below then we know
   the scalar is directly selected.  If not it's in a gop canvas.  (This doesn't
   yet handle nested GOPs, unfortunately.)
*/
static void scalar_displace_withtag(t_gobj *z, t_glist *glist, int dx, int dy)
{
    //fprintf(stderr,"scalar_displace_withtag %lx %d %d\n", (t_int)z, dx, dy);
    t_scalar *x = (t_scalar *)z;
    t_symbol *templatesym = x->sc_template;
    t_template *template = template_findbyname(templatesym);
    t_symbol *zz;
    t_atom at[3];
    t_gpointer gp;
    int xonset, yonset, xtype, ytype, gotx, goty;
    int bx1, bx2, by1, by2;
    t_float basex = 0, basey = 0;
    if (!template)
    {
        error("scalar: couldn't find template %s", templatesym->s_name);
        return;
    }
    gotx = template_find_field(template, gensym("x"), &xonset, &xtype, &zz);
    if ((gotx && (xtype != DT_FLOAT)) || x->sc_selected != glist)
        gotx = 0;
    goty = template_find_field(template, gensym("y"), &yonset, &ytype, &zz);
    if ((goty && (ytype != DT_FLOAT)) || x->sc_selected != glist)
        goty = 0;
    if (gotx)
    {
        *(t_float *)(((char *)(x->sc_vec)) + xonset) +=
            dx * (glist_pixelstox(glist, 1) - glist_pixelstox(glist, 0));
        x->sc_x1 += dx;
        x->sc_x2 += dx;
    }
    if (goty)
    {
        *(t_float *)(((char *)(x->sc_vec)) + yonset) +=
            dy * (glist_pixelstoy(glist, 1) - glist_pixelstoy(glist, 0));
        x->sc_y1 += dy;
        x->sc_y2 += dy;
    }
    scalar_getbasexy(x, &basex, &basey);
    gpointer_init(&gp);
    gpointer_setglist(&gp, glist, x);
    SETPOINTER(&at[0], &gp);
    SETFLOAT(&at[1], (t_float)dx);
    SETFLOAT(&at[2], (t_float)dy);
    template_notify(template, gensym("displace"), 2, at);

    /* this is a hack to make sure the bbox gets drawn in
       the right location.  If the scalar is selected
       then it's possible that a "displace" message
       from the [struct] will trigger a redraw of the
       bbox. So we don't update the cached bbox until
       after that redraw, so we can move the bbox below.
    */
    sys_vgui(".x%lx.c coords {select%lx} %d %d %d %d\n", glist, x,
        x->sc_x1 - 1, x->sc_y1 - 1, x->sc_x2 + 1, x->sc_y2 + 1);

    t_float xscale = glist_xtopixels(x->sc_selected, 1) -
        glist_xtopixels(x->sc_selected, 0);
    t_float yscale = glist_ytopixels(x->sc_selected, 1) -
        glist_ytopixels(x->sc_selected, 0);

    sys_vgui(".x%lx.c itemconfigure {.scalar%lx} "
             "-matrix { {%g %g} {%g %g} {%d %d} }\n",
        glist_getcanvas(glist), x->sc_vec, xscale, 0.0, 0.0, yscale,
        (int)glist_xtopixels(x->sc_selected, basex) +
            (x->sc_selected == glist ? 0 : dx),
        (int)glist_ytopixels(x->sc_selected, basey) +
            (x->sc_selected == glist ? 0 : dy));

    /* This awful hack is here because plot_vis is used by both ds arrays
       and garrays.  Unlike the other drawing commands, plot_vis still does
       all the gop scaling and basex/basey calculations manually.  So
       currently the trace does not use the scalar group as its "-parent".

       Once all the calculations inside plot_vis, array_doclick, and
       array_motion remove the glist_[xy]topixels and basex/y stuff, plot_vis
       can use the scalar "-parent" and this awful hack can be removed.

       Garrays follow their Pd-l2ork's standard select and displace_withtag
       behavior, so we don't use the hack for them.  (Non-garray scalars
       should follow that behavior too, but cannot atm for the reason given
       in the comment above scalar_select...)

       Apparently this is no longer needed, so it is commented out.  Once
       we test it we should be able to delete it for good... */

    /*
    if (template->t_sym != gensym("_float_array"))
    {
        t_gobj *y;
        t_canvas *templatecanvas = template_findcanvas(template);
        for (y = templatecanvas->gl_list; y; y = y->g_next)
            {
                t_parentwidgetbehavior *wb = pd_getparentwidget(&y->g_pd);
                if (!wb) continue;
                (*wb->w_parentdisplacefn)(y, glist, x->sc_vec, template,
                    basex, basey, dx, dy);
            }
    }
    */

    //scalar_redraw(x, glist);
}

static void scalar_activate(t_gobj *z, t_glist *owner, int state)
{
    /* post("scalar_activate %d", state); */
    /* later */
}

static void scalar_delete(t_gobj *z, t_glist *glist)
{
    /* nothing to do */
}

/* At present, scalars have a three-level hierarchy in tkpath,
   with two levels accessible by the user from within Pd:
   scalar - tkpath group with matrix derived from x/y fields,
     |      gop basexy, and gop scaling values. This group is
     |      not configurable by the user. This means that the
     |      [draw group] below can ignore basexy and gop junk
     |      when computing the transform matrix.
     v
   dgroup - user-facing group which is the parent for all the
     |      scalar's drawing commands. Its matrix and options
     |      can be accessed from the [draw group] object (one
     |      per templatecanvas).
     v
   draw   - the actual drawing command: rectangle, path, etc.
            Each has its own matrix and options which can set
            with messages to the corresponding [draw] object.

   The tag "blankscalar" is for scalars that don't have a visual
   representation, but maybe this can just be merged with "scalar"
*/
static void scalar_vis(t_gobj *z, t_glist *owner, int vis)
{
    //fprintf(stderr,"scalar_vis %d\n", vis);
    t_scalar *x = (t_scalar *)z;

    x->sc_bboxcache = 0;

    t_template *template = template_findbyname(x->sc_template);
    t_canvas *templatecanvas = template_findcanvas(template);
    t_gobj *y;
    t_float basex, basey;
    scalar_getbasexy(x, &basex, &basey);
        /* if we don't know how to draw it, make a small rectangle */
    if (!templatecanvas)
    {
        if (vis)
        {
            int x1 = glist_xtopixels(owner, basex);
            int y1 = glist_ytopixels(owner, basey);
            sys_vgui(".x%lx.c create prect %d %d %d %d "
                     "-tags {blankscalar%lx}\n",
                glist_getcanvas(owner), x1-1, y1-1, x1+1, y1+1, x);
        }
        else sys_vgui(".x%lx.c delete blankscalar%lx\n",
            glist_getcanvas(owner), x);
        return;
    }
    //else sys_vgui(".x%lx.c delete blankscalar%lx\n",
    //    glist_getcanvas(owner), x);

    if (vis)
    {
        t_float xscale = glist_xtopixels(owner, 1) - glist_xtopixels(owner, 0);
        t_float yscale = glist_ytopixels(owner, 1) - glist_ytopixels(owner, 0);
        /* we could use the tag .template%lx for easy access from
           the draw_class, but that's not necessary at this point */
        sys_vgui(".x%lx.c create group -tags {.scalar%lx} "
            "-matrix { {%g %g} {%g %g} {%d %d} }\n",
            glist_getcanvas(owner), x->sc_vec,
            xscale, 0.0, 0.0, yscale, (int)glist_xtopixels(owner, basex),
            (int)glist_ytopixels(owner, basey)
            );
        sys_vgui(".x%lx.c create group -tags {.dgroup%lx} "
                 "-parent {.scalar%lx}\n",
            glist_getcanvas(owner), x->sc_vec, x->sc_vec);
        sys_vgui("pdtk_bind_scalar_mouseover "
                 ".x%lx.c .x%lx.x%lx.template%lx {.x%lx}\n",
            glist_getcanvas(owner), glist_getcanvas(owner),
            owner, x->sc_vec, x);
    }

    for (y = templatecanvas->gl_list; y; y = y->g_next)
    {
        t_parentwidgetbehavior *wb = pd_getparentwidget(&y->g_pd);
        if (!wb) continue;
        (*wb->w_parentvisfn)(y, owner, x, x->sc_vec, template,
            basex, basey, vis);
    }
    if (!vis)
        sys_vgui(".x%lx.c delete .scalar%lx\n", glist_getcanvas(owner),
            x->sc_vec);

    sys_unqueuegui(x);
    if (glist_isselected(owner, &x->sc_gobj))
    {
        //scalar_select(z, owner, 1);
        scalar_drawselectrect(x, owner, 0);
        scalar_drawselectrect(x, owner, 1);
    }
}

static void scalar_doredraw(t_gobj *client, t_glist *glist)
{
    scalar_vis(client, glist, 0);
    scalar_vis(client, glist, 1);
    if (glist_isselected(glist_getcanvas(glist), (t_gobj *)glist))
    {
        //fprintf(stderr,"yes\n");
        sys_vgui("pdtk_select_all_gop_widgets .x%lx %lx %d\n",
            glist_getcanvas(glist), glist, 1);
    }
    sys_vgui("pdtk_canvas_getscroll .x%lx.c\n", glist_getcanvas(glist));
}

void scalar_redraw(t_scalar *x, t_glist *glist)
{
    if (glist_isvisible(glist))
        scalar_doredraw((t_gobj *)x, glist);
        //sys_queuegui(x, glist, scalar_doredraw);
}

int scalar_doclick(t_word *data, t_template *template, t_scalar *sc,
    t_array *ap, struct _glist *owner,
    t_float xloc, t_float yloc, int xpix, int ypix,
    int shift, int alt, int dbl, int doit)
{
    int hit = 0;
    t_canvas *templatecanvas = template_findcanvas(template);
    t_gobj *y;
    t_atom at[2];
    t_float basex = template_getfloat(template, gensym("x"), data, 0);
    t_float basey = template_getfloat(template, gensym("y"), data, 0);
    //fprintf(stderr,"=================scalar_doclick %f %f %f %f %d\n",
    //    basex, basey, xloc, yloc, doit);

    SETFLOAT(at, basex + xloc);
    SETFLOAT(at+1, basey + yloc);
    if (doit)
    {
        //fprintf(stderr,"    doit\n");
        template_notifyforscalar(template, owner, 
            sc, gensym("click"), 2, at);
    }

    // if we are nested ignore xloc and yloc, otherwise
    // nested objects get their hitbox miscalculated
    if (xloc != 0.0 || yloc != 0.0)
    {
        //fprintf(stderr,"ignoring\n");
        basex = 0.0;
        basey = 0.0;
    }

    for (y = templatecanvas->gl_list; y; y = y->g_next)
    {
        //fprintf(stderr,"looking for template... %f %f %f %f %lx %lx\n",
        //    basex, basey, xloc, yloc, (t_int)owner, (t_int)data);
        t_parentwidgetbehavior *wb = pd_getparentwidget(&y->g_pd);
        if (!wb) continue;
        if (hit = (*wb->w_parentclickfn)(y, owner,
            data, template, sc, ap, basex + xloc, basey + yloc,
            xpix, ypix, shift, alt, dbl, doit))
        {
                //fprintf(stderr,"    ...got it %f %f\n",
                //    basex + xloc, basey + yloc);
                return (hit);
        }
    }
    return (0);
}

static int scalar_click(t_gobj *z, struct _glist *owner,
    int xpix, int ypix, int shift, int alt, int dbl, int doit)
{
    //fprintf(stderr,"scalar_click %d %d\n", xpix, ypix);
    t_scalar *x = (t_scalar *)z;

    x->sc_bboxcache = 0;

    t_template *template = template_findbyname(x->sc_template);

    return (scalar_doclick(x->sc_vec, template, x, 0,
        owner, 0, 0, xpix, ypix, shift, alt, dbl, doit));
}

void canvas_writescalar(t_symbol *templatesym, t_word *w, t_binbuf *b,
    int amarrayelement);

static void scalar_save(t_gobj *z, t_binbuf *b)
{
    t_scalar *x = (t_scalar *)z;
    t_binbuf *b2 = binbuf_new();
    t_atom a, *argv;
    int i, argc;
    canvas_writescalar(x->sc_template, x->sc_vec, b2, 0);
    binbuf_addv(b, "ss", &s__X, gensym("scalar"));
    binbuf_addbinbuf(b, b2);
    binbuf_addsemi(b);
    binbuf_free(b2);
}

static void scalar_properties(t_gobj *z, struct _glist *owner)
{
    t_scalar *x = (t_scalar *)z;
    char *buf, buf2[80];
    int bufsize;
    t_binbuf *b;
    glist_noselect(owner);
    glist_select(owner, z);
    b = glist_writetobinbuf(owner, 0);
    binbuf_gettext(b, &buf, &bufsize);
    binbuf_free(b);
    buf = t_resizebytes(buf, bufsize, bufsize+1);
    buf[bufsize] = 0;
    sprintf(buf2, "pdtk_data_dialog %%s {");
    gfxstub_new((t_pd *)owner, x, buf2);
    sys_gui(buf);
    sys_gui("}\n");
    t_freebytes(buf, bufsize+1);
}

static t_widgetbehavior scalar_widgetbehavior =
{
    scalar_getrect,
    scalar_displace,
    scalar_select,
    scalar_activate,
    scalar_delete,
    scalar_vis,
    scalar_click,
    scalar_displace_withtag,
};

static void scalar_free(t_scalar *x)
{
    int i;
    t_dataslot *datatypes, *dt;
    t_symbol *templatesym = x->sc_template;
    t_template *template = template_findbyname(templatesym);
    if (!template)
    {
        error("scalar: couldn't find template %s", templatesym->s_name);
        return;
    }
    word_free(x->sc_vec, template);
    char buf[50];
    sprintf(buf, ".x%lx", (long unsigned int)x);
    pd_unbind(&x->sc_gobj.g_pd, gensym(buf));
    gfxstub_deleteforkey(x);
        /* the "size" field in the class is zero, so Pd doesn't try to free
        us automatically (see pd_free()) */
    freebytes(x, sizeof(t_scalar) + (template->t_n - 1) * sizeof(*x->sc_vec));
}

/* ----------------- setup function ------------------- */

void g_scalar_setup(void)
{
    scalar_class = class_new(gensym("scalar"), 0, (t_method)scalar_free, 0,
        CLASS_GOBJ, 0);
    class_addmethod(scalar_class, (t_method)scalar_mouseover,
        gensym("mouseover"), A_FLOAT, A_NULL);
    class_setwidget(scalar_class, &scalar_widgetbehavior);
    class_setsavefn(scalar_class, scalar_save);
    class_setpropertiesfn(scalar_class, scalar_properties);
}