static dim do_remove
( mclx* mx
, dim N_remove
, dim* offsets /* size N_COLS(mx) */
, dim N_edge
, dim random_ignore
)
{ dim n_remove = 0
; while (n_remove < N_remove)
{ unsigned long r = (unsigned long) random()
; dim e, *ep
; ofs xo, yo
; long x
; mclv* vecx
; if (r >= random_ignore)
continue
; e = r % N_edge
; if (!(ep = mcxBsearchFloor(&e, offsets, N_COLS(mx), sizeof e, dimCmp)))
mcxDie(1, me, "edge %ld not found (max %ld)", (long) e, (long) N_edge)
; xo = ep - offsets
;if(DEBUG)fprintf(stderr, "have e=%d o=%d\n", (int) e, (int) xo)
; x = mx->dom_cols->ivps[xo].idx
; vecx = mx->cols+xo;
; yo = e - offsets[xo]
; if (yo >= vecx->n_ivps) /* note: mixed sign comparison */
mcxDie
( 1
, me
, "vector %ld has not enough entries: %ld < %ld"
, (long) vecx->vid
, (long) vecx->n_ivps
, (long) yo
)
; if (!vecx->ivps[yo].val)
continue
; vecx->ivps[yo].val = 0.0
;if(DEBUG)fprintf(stderr, "remove [%d] %ld %ld\n", (int) n_remove, (long) x, (long) vecx->ivps[yo].idx)
; n_remove++
; }
return n_remove
; }
int main
( int argc
, const char* argv[]
)
{ const char* fn_input = argc > 1 ? argv[1] : ""
; mclAlgParam* mlp = NULL
; const char* me = "mcl"
; mcxstatus status = STATUS_OK
; srandom(mcxSeed(315))
; signal(SIGALRM, mclSigCatch)
; if (signal(SIGUSR1, mcxLogSig) == SIG_ERR)
mcxErr(me, "cannot catch SIGUSR1!")
; mcxLogLevel =
MCX_LOG_AGGR | MCX_LOG_MODULE | MCX_LOG_IO | MCX_LOG_GAUGE | MCX_LOG_WARN
; mclx_app_init(stderr)
; if (argc < 2)
{ mcxTell
( me
, "usage: mcl <-|file name> [options],"
" do 'mcl -h' or 'man mcl' for help"
)
; mcxExit(0)
; }
status
= mclAlgInterface
(&mlp, (char**) (argv+2), argc-2, fn_input, NULL, ALG_DO_IO)
; if (status == ALG_INIT_DONE)
return 0
; else if (status)
mcxDie(STATUS_FAIL, me, "no tango")
; if ((status = mclAlgorithm(mlp)) == STATUS_FAIL)
mcxDie(STATUS_FAIL, me, "failed")
; if (mlp->n_assimilated)
mcxLog(MCX_LOG_MODULE, me, "%lu nodes will assimilate", (ulong) mlp->n_assimilated)
; if (mlp->mx_start)
mcxLog(MCX_LOG_MODULE, me, "cached matrix with %lu columns", (ulong) N_COLS(mlp->mx_start))
; mclAlgParamFree(&mlp, TRUE)
; helpful_reminder()
; return STATUS_OK
; }
dim get_n_sort_allvals
( const mclx* mx
, pval* allvals
, dim noe
, double* sum_vals
, mcxbool asc
)
{ dim n_allvals = 0
; double s = 0.0
; dim j
; for (j=0;j<N_COLS(mx);j++)
{ mclv* vec = mx->cols+j
; dim k
; if (n_allvals + vec->n_ivps > noe)
mcxDie(1, me, "panic/impossible: not enough memory")
; for (k=0;k<vec->n_ivps;k++)
allvals[n_allvals+k] = vec->ivps[k].val
; n_allvals += vec->n_ivps
; s += mclvSum(vec)
; }
if (asc)
qsort(allvals, n_allvals, sizeof(pval), valcmpasc)
; else
qsort(allvals, n_allvals, sizeof(pval), valcmpdesc)
; sum_vals[0] = s
; return n_allvals
; }
static mcxstatus convertMain
( int argc_unused cpl__unused
, const char* argv[]
)
{ mclMatrix* mx
; mclxCat st
; xfin = mcxIOnew(argv[0], "r")
; xfout = mcxIOnew(argv[1], "w")
; mclxCatInit(&st)
; if (main_mode == 'l')
{ if (mclxCatRead(xfin, &st, catmax, NULL, NULL, 0))
mcxDie(1, me, "failure is, if not an option, the result after all")
; mclxCatWrite(xfout, &st, MCLXIO_VALUE_GETENV, EXIT_ON_FAIL)
; }
else if (main_mode == 'f')
{ int format
; mx = mclxRead(xfin, EXIT_ON_FAIL)
; format = mclxIOformat(xfin)
; if (!test_read)
{ mcxIOopen(xfout, EXIT_ON_FAIL)
; if (format == 'a')
mclxbWrite(mx, xfout, EXIT_ON_FAIL)
; else
mclxaWrite(mx, xfout, MCLXIO_VALUE_GETENV, EXIT_ON_FAIL)
; }
}
else
{ mcxbits bits
= main_mode == 'c'
? MCLX_REQUIRE_DOMTREE | MCLX_CATREAD_CLUSTERSTACK
: main_mode == 's'
? MCLX_REQUIRE_DOMSTACK | MCLX_CATREAD_CLUSTERTREE
: 0
; mcxIOopen(xfout, EXIT_ON_FAIL)
; if (mclxCatRead(xfin, &st, catmax, NULL, NULL, bits))
mcxDie(1, me, "failure is, if not an option, the result after all")
; mclxCatWrite(xfout, &st, MCLXIO_VALUE_NONE, EXIT_ON_FAIL)
; }
return 0
; }
void* mclxComposeThreadData
( mclxComposeHelper* ch
, int i_thread
)
{ if (i_thread > ch->n_jobs)
mcxDie(1, "compose", "fatal: thread ID error (%d asked, %d max)", (int) i_thread, (int) ch->n_jobs)
; return ch->iovs[i_thread]
; }
int main
( int argc
, const char* argv[]
)
{ mclx* mx_score, *mx_start, *mx_end
; mcxIO* xfs, *xfl, *xfr
; dim i, j
; if (argc != 4)
mcxDie(1, me, "need score start end matrices")
; xfs = mcxIOnew(argv[1], "r")
; xfl = mcxIOnew(argv[2], "r")
; xfr = mcxIOnew(argv[3], "r")
; mx_score = mclxRead(xfs, EXIT_ON_FAIL)
; mx_start = mclxRead(xfl, EXIT_ON_FAIL)
; mx_end = mclxRead(xfr, EXIT_ON_FAIL)
; fprintf
( stdout
, "%10s%10s%10s%10s%10s%10s%10s%10s%10s%10s%10s%10s\n"
, "vid1"
, "vid2"
, "diff1"
, "diff2"
, "meet"
, "score"
, "mean"
, "csn"
, "euclid"
, "frac"
, "re1"
, "re2"
)
; for (i=0;i< N_COLS(mx_score);i++)
{ for (j=0;j<N_COLS(mx_score);j++)
{ double s
; s =
get_score
( mx_score->cols+i
, mx_score->cols+j
, mx_start->cols+i
, mx_start->cols+j
, mx_end->cols+i
, mx_end->cols+j
)
; }
}
return 0
; }
static mclx* get_base
( const char* fn_input
, mclx* mx_input
, const char* b1opts
, const char* b2opts
)
{ mclAlgParam* mlp = NULL
; mcxstatus status = STATUS_FAIL
; mclx* mxbase = NULL
; get_interface
( &mlp
, fn_input
, b2opts ? b2opts : b1opts
, NULL
, mx_input
, b2opts ? ALG_CACHE_EXPANDED : ALG_CACHE_START
, EXIT_ON_FAIL
)
; if (b2opts)
{ mlp->mpp->mainLoopLength = 1
; mlp->mpp->mainInflation = 1.0
; mlp->expand_only = TRUE
; mcxLog(MCX_LOG_APP, me, "computing expanded matrix")
; if ((status = mclAlgorithm(mlp)) == STATUS_FAIL)
{ mcxErr(me, "failed")
; mcxExit(1)
; }
mcxLog(MCX_LOG_APP, me, "done computing expanded matrix")
; mxbase = mclAlgParamRelease(mlp, mlp->mx_expanded)
; }
else
{ if (mclAlgorithmStart(mlp, FALSE))
mcxDie(1, me, "-b1 option start-up failed")
; mxbase = mclAlgParamRelease(mlp, mlp->mx_start)
; start_col_sums_g = mclvClone(mlp->mx_start_sums)
; }
mclAlgParamFree(&mlp, TRUE)
; return mxbase
; }
mcxstatus mcxIOclose
( mcxIO* xf
)
{ fflush(xf->fp)
;if (!strcmp(xf->fn->str, "-") && !strcmp(xf->mode, "w") && !xf->stdio)
mcxDie(1, "tst", "should not happen")
; if (xf->fp && !xf->stdio)
{ fclose(xf->fp)
; xf->fp = NULL
; }
else if (xf->fp && xf->stdio)
{ int fe = ferror(xf->fp) /* fixme why not in branch above? */
; if (fe)
mcxErr("mcxIOclose", "error [%d] for [%s] stdio", fe, xf->mode)
, perror("mcxIOclose")
; if (xf->ateof || feof(xf->fp))
clearerr(xf->fp)
; }
/* fixme contract with usr_reset */
return mcxIOreset(xf)
; }
void static do_the_shuffle
( mclx* mx
, dim N_shuffle
, dim* offsets /* size N_COLS(mx) */
, dim N_edge
, dim random_ignore
)
{ dim n_shuffle = 0
; while (n_shuffle < N_shuffle)
{ unsigned long rx = (unsigned long) random()
; unsigned long ry = (unsigned long) random()
; mclp* ivpll, *ivplr, *ivprl, *ivprr
; dim edge_x, edge_y, *edge_px, *edge_py
; ofs xro, yro, xlo, ylo = -1, vxo, vyo
; long xl, xr, yl, yr
; mclv* vecxl, *vecxr, *vecyl, *vecyr
; double xlval, xrval, ylval, yrval
; if (rx >= random_ignore || ry >= random_ignore)
continue
; edge_x = rx % N_edge /* fixme probably not optimal */
; edge_y = ry % N_edge /* fixme probably not optimal */
; if (!(edge_px = mcxBsearchFloor(&edge_x, offsets, N_COLS(mx), sizeof edge_x, dimCmp)))
mcxDie(1, me, "edge %ld not found (max %ld)", (long) edge_x, (long) N_edge)
; if (!(edge_py = mcxBsearchFloor(&edge_y, offsets, N_COLS(mx), sizeof edge_y, dimCmp)))
mcxDie(1, me, "edge %ld not found (max %ld)", (long) edge_y, (long) N_edge)
; vxo = edge_px - offsets
; xl = mx->dom_cols->ivps[vxo].idx
; vecxl = mx->cols+vxo
; xro = edge_x - offsets[vxo]
; vyo = edge_py - offsets
; yl = mx->dom_cols->ivps[vyo].idx
; vecyl = mx->cols+vyo
; yro = edge_y - offsets[vyo]
/* Offset computation gone haywire */
; if (xro >= vecxl->n_ivps || yro >= vecyl->n_ivps) /* note: mixed sign comparison */
mcxDie(1, me, "paradox 1 in %ld or %ld", xl, yl)
; xr = vecxl->ivps[xro].idx
; yr = vecyl->ivps[yro].idx
; xrval = vecxl->ivps[xro].val
; yrval = vecyl->ivps[yro].val
/* Impossible, should have graph */
; vecxr = mclxGetVector(mx, xr, EXIT_ON_FAIL, NULL)
; vecyr = mclxGetVector(mx, yr, EXIT_ON_FAIL, NULL)
/* check that we have four different nodes
* loops are not present so no need to check those
*/
; if (xl == yl || xl == yr || xr == yl || xr == yr)
continue
; if
( (0 > (xlo = mclvGetIvpOffset(vecxr, xl, -1)))
|| (0 > (ylo = mclvGetIvpOffset(vecyr, yl, -1)))
)
mcxDie
( 1
, me
, "symmetry violation 1"
" %ld not found in %ld/%ld OR %ld not found in %ld/%ld"
, (long) xl, (long) vecxr->vid, (long) vecxr->n_ivps
, (long) yl, (long) vecyr->vid, (long) vecyr->n_ivps
)
/* Now: xl yl : ivpll
* xl yr : ivplr
* xr yl : ivprl
* xr yr : ivprr
*/
; xlval = vecxr->ivps[xlo].val
; ylval = vecyr->ivps[ylo].val
; ivpll = mclvGetIvp(vecxl, yl, NULL)
; ivplr = mclvGetIvp(vecxl, yr, NULL)
; ivprl = mclvGetIvp(vecxr, yl, NULL)
; ivprr = mclvGetIvp(vecxr, yr, NULL)
; if
( (ivpll && !mclvGetIvp(vecyl, xl, NULL))
|| (ivplr && !mclvGetIvp(vecyr, xl, NULL))
|| (ivprl && !mclvGetIvp(vecyl, xr, NULL))
|| (ivprr && !mclvGetIvp(vecyr, xr, NULL))
)
mcxDie(1, me, "symmetry violation 2")
; if ((ivpll && ivplr) || (ivprl && ivprr))
continue
; { if (!ivpll && !ivprr)
{ /* vecxl <-> xr becomes vecxl <-> yl
* vecxr <-> xl becomes vecxr <-> yr
* vecyl <-> yr becomes vecyl <-> xl
* vecyr <-> yl becomes vecyr <-> xr
*/
; if
( mclvReplaceIdx(vecxl, xro, yl, xrval)
|| mclvReplaceIdx(vecyl, yro, xl, xrval)
|| mclvReplaceIdx(vecxr, xlo, yr, ylval)
|| mclvReplaceIdx(vecyr, ylo, xr, ylval)
)
mcxDie(1, me, "parallel replacement failure\n")
#if DEBUG
;fprintf(stderr, "parallel edge change remove %d-%d %d-%d add %d-%d %d-%d\n",
vecxl->vid, xr, vecyr->vid, yl, vecxl->vid, yl, vecyr->vid, xr)
#endif
; }
else if (!ivplr && !ivprl)
{ /* vecxl -> xr becomes vecxl <-> yr
* vecxr -> xl becomes vecxr <-> yl
* vecyl -> yr becomes vecyl <-> xr
* vecyr -> yl becomes vecyr <-> xl
*/
if
( mclvReplaceIdx(vecxl, xro, yr, xrval)
|| mclvReplaceIdx(vecyr, ylo, xl, xlval)
|| mclvReplaceIdx(vecxr, xlo, yl, yrval)
|| mclvReplaceIdx(vecyl, yro, xr, yrval)
)
mcxDie(1, me, "cross replacement failure\n")
#if DEBUG
;fprintf(stderr, "cross edge change remove %d-%d %d-%d add %d-%d %d-%d\n",
vecxl->vid, xr, vecyl->vid, yr, vecxl->vid, yr, vecyl->vid, xr)
#endif
; }
}
n_shuffle++
; }
}
int main
( int argc
, const char* argv[]
)
{ mcxIO
*xfcl = NULL
, *xfctrl = NULL
, *xfcoarse= NULL
, *xfbase = NULL
, *xfcone = NULL
, *xfstack = NULL
; mclx* mxbase, *cl, *cl_coarse, *clprev, *clctrl = NULL
; mcxTing* shared = mcxTingNew("-I 4 -overlap split")
; mcxbool root = TRUE
; mcxbool have_bootstrap = FALSE
; const char* plexprefix = NULL
; const char* stem = "mcl"
; mcxbool same = FALSE
; mcxbool plex = TRUE
; mcxbool add_transpose = FALSE
; const char* b2opts = NULL
; const char* b1opts = NULL
; mcxbits write_modes = 0
; mclAlgParam* mlp = NULL
; mcxstatus status = STATUS_OK
; mcxstatus parse_status = STATUS_OK
; int multiplex_idx = 1
; int N = 0
; int n_ite = 0
; dim n_components = 0, n_cls = 0
; int a = 1, i= 0
; int n_arg_read = 0
; int delta = 0
; mcxOption* opts, *opt
; mcxTing* cline = mcxOptArgLine(argv+1, argc-1, '\'')
; mclgTF* transform = NULL
; mcxTing* transform_spec = NULL
; double iaf = 0.84
; mclx_app_init(stderr)
; if (0)
mcxLogLevel =
MCX_LOG_AGGR | MCX_LOG_MODULE | MCX_LOG_IO | MCX_LOG_GAUGE | MCX_LOG_WARN
; else
mcxLogLevelSetByString("xf4g1")
; mcxOptAnchorSortById(options, sizeof(options)/sizeof(mcxOptAnchor) -1)
; if (argc == 2 && argv[1][0] == '-' && mcxOptIsInfo(argv[1], options))
delta = 1
; else if (argc < 2)
{ help(options, shared)
; exit(0)
; }
opts = mcxOptExhaust
(options, (char**) argv, argc, 2-delta, &n_arg_read, &parse_status)
; if (parse_status != STATUS_OK)
{ mcxErr(me, "initialization failed")
; exit(1)
; }
; for (opt=opts;opt->anch;opt++)
{ mcxOptAnchor* anch = opt->anch
; switch(anch->id)
{ case MY_OPT_HELP
: help(options, shared)
; exit(0)
;
case MY_OPT_APROPOS
: help(options, shared)
; exit(0)
; break
;
case MY_OPT_NMAX
: N = atoi(opt->val)
; break
;
case MY_OPT_Z
: help(NULL, shared)
; exit(0)
; break
;
case MY_OPT_SHARED
: mcxTingPrintAfter(shared, " %s", opt->val)
; break
;
case MY_OPT_TRANSFORM
: transform_spec = mcxTingNew(opt->val)
; break
;
case MY_OPT_B1
: b1opts = opt->val
; break
;
case MY_OPT_B2
: b2opts = opt->val
; break
;
case ALG_OPT_SETENV
: mcxSetenv(opt->val)
; break
;
case ALG_OPT_QUIET
: mcxLogLevelSetByString(opt->val)
; break
;
case MY_OPT_HDP
: hdp_g = atof(opt->val)
; break
;
case MY_OPT_ADDTP
: add_transpose = TRUE
; break
;
case MY_OPT_ANNOT /* only used in command-line copying */
: break
;
case MY_OPT_IAF
: iaf = atof(opt->val) / 100
; break
;
case MY_OPT_WRITE
: if (strstr(opt->val, "stack"))
write_modes |= OUTPUT_STACK
; if (strstr(opt->val, "cone"))
write_modes |= OUTPUT_CONE
; if (strstr(opt->val, "levels"))
write_modes |= OUTPUT_STEPS
; if (strstr(opt->val, "coarse"))
write_modes |= OUTPUT_COARSE
; if (strstr(opt->val, "base"))
write_modes |= OUTPUT_BASE
; break
;
case MY_OPT_BASENAME
: xfbase = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_COARSE
: xfcoarse = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_CONE
: xfcone = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_ROOT
: root = strchr("1yY", (u8) opt->val[0]) ? TRUE : FALSE
; break
;
case MY_OPT_STACK
: xfstack = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_STEM
: stem = opt->val
; break
;
case MY_OPT_MULTIPLEX
: plex = strchr("yY1", (unsigned char) opt->val[0]) ? TRUE : FALSE
; break
;
case MY_OPT_DISPATCH
: dispatch_g = TRUE
; break
;
case MY_OPT_INTEGRATE
: integrate_g = TRUE
; break
;
case MY_OPT_CONTRACT
: break
;
case MY_OPT_SUBCLUSTERX
: subclusterx_g = TRUE, subcluster_g = TRUE
; break
;
case MY_OPT_SUBCLUSTER
: subcluster_g = TRUE
; break
;
case MY_OPT_CONTROL
: xfctrl = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_CL
: xfcl = mcxIOnew(opt->val, "r")
; have_bootstrap = TRUE
; break
;
case MY_OPT_VERSION
: app_report_version(me)
; exit(0)
;
default
: mcxExit(1)
;
}
}
mcxOptFree(&opts)
; a = 2 + n_arg_read
; if (a < argc)
{ if (strcmp(argv[a], "--"))
mcxDie
( 1
, me
, "trailing %s options require standalone '--' separator (found %s)"
, integrate_g ? "integrate" : "mcl"
, argv[a]
)
; a++
; }
if (subcluster_g + dispatch_g + integrate_g > 1)
mcxDie(1, me, "too many modes!")
; if (N && N < argc-a)
mcxErr(me, "-n argument leaves spurious option specifications")
; srandom(mcxSeed(89315))
; signal(SIGALRM, mclSigCatch)
; if (dispatch_g)
plexprefix = "dis"
; else if (!write_modes || (write_modes & OUTPUT_STEPS))
plexprefix = stem
; { mcxTing* tg = mcxTingEmpty(NULL, 30)
; if ((write_modes & OUTPUT_COARSE) && !xfcoarse)
mcxTingPrint(tg, "%s.%s", stem, "coarse")
, xfcoarse = mcxIOnew(tg->str, "w")
; if ((write_modes & OUTPUT_BASE) && !xfbase)
mcxTingPrint(tg, "%s.%s", stem, "base")
, xfbase = mcxIOnew(tg->str, "w")
; if
( (!write_modes || (write_modes & OUTPUT_CONE))
&& !xfcone
)
{ mcxTingPrint(tg, "%s.%s", stem, "cone")
; xfcone = mcxIOnew(tg->str, "w")
; mcxIOopen(xfcone, EXIT_ON_FAIL)
; fprintf(xfcone->fp, "# %s %s\n", argv[0], cline->str)
; }
if ((write_modes & OUTPUT_STACK) && !xfstack)
{ mcxTingPrint(tg, "%s.%s", stem, "stack")
; xfstack = mcxIOnew(tg->str, "w")
; mcxIOopen(xfstack, EXIT_ON_FAIL)
; fprintf(xfstack->fp, "# %s %s\n", argv[0], cline->str)
; }
mcxTingFree(&tg)
; }
if (integrate_g)
{ for (i=a;i<argc;i++)
{ mcxIO* xf = mcxIOnew(argv[i], "r")
; mclx* cl = mclxRead(xf, EXIT_ON_FAIL)
; mclxCatPush(&stck_g, cl, NULL, NULL, mclxCBdomStack, NULL, "dummy-integrate", n_cls++)
; }
integrate_results(&stck_g)
; if (xfstack)
mclxCatWrite(xfstack, &stck_g, MCLXIO_VALUE_NONE, RETURN_ON_FAIL)
; if (xfcone)
mclxCatConify(&stck_g)
, mclxCatWrite(xfcone, &stck_g, MCLXIO_VALUE_NONE, RETURN_ON_FAIL)
; return 0
; }
for (i=a;i<argc;i++)
{ if (get_interface(NULL, argv[1], shared->str, argv[i], NULL, 0, RETURN_ON_FAIL))
mcxDie(1, me, "error while testing mcl options viability (%s)", argv[i])
; }
mcxLog(MCX_LOG_APP, me, "pid %ld", (long) getpid())
/* make sure clusters align with this cluster
* status: does not seem promising.
*/
; if (xfctrl)
clctrl = mclxRead(xfctrl, EXIT_ON_FAIL)
;
/*
* Below: compute cl and mxbase.
*/
; if (xfcl)
{ cl = mclxRead(xfcl, EXIT_ON_FAIL)
; write_clustering
(cl, NULL, xfcone, xfstack, plexprefix, multiplex_idx++, NULL)
; if (subcluster_g || dispatch_g)
mclxCatPush(&stck_g, cl, NULL, NULL, mclxCBdomStack, NULL, "dummy-mclcm", n_cls++)
; mcxIOfree(&xfcl)
; if (!b1opts && !b2opts)
b1opts = ""
; mxbase = get_base(argv[1], NULL, b1opts, b2opts)
; }
else
{ mcxbits CACHE = b1opts || b2opts
? ALG_CACHE_INPUT /* cache, transform later */
: ALG_CACHE_START
; get_interface
( &mlp
, argv[1]
, shared->str
, a < argc ? argv[a] : NULL
, NULL
, CACHE
, EXIT_ON_FAIL
)
; if (a < argc)
a++
; if ((status = mclAlgorithm(mlp)) == STATUS_FAIL)
{ mcxErr(me, "failed at initial run")
; exit(1)
; }
cl_coarse = mclAlgParamRelease(mlp, mlp->cl_result)
; cl_coarse = control_test(cl_coarse, clctrl)
; write_clustering
(cl_coarse, NULL, xfcone, xfstack, plexprefix, multiplex_idx++, mlp)
; if (subcluster_g || dispatch_g)
mclxCatPush(&stck_g, cl_coarse, NULL, NULL, mclxCBdomStack, NULL, "dummy-mclcm", n_cls++)
; cl = cl_coarse
; n_ite++
; if (b1opts || b2opts)
{ mclx* mx_input = mclAlgParamRelease(mlp, mlp->mx_input)
; mxbase = get_base(NULL, mx_input, b1opts, b2opts)
/* ^ get_base frees mx_input */
; }
else
mxbase = mclAlgParamRelease(mlp, mlp->mx_start)
; }
clprev = cl
; mclAlgParamFree(&mlp, TRUE)
; if (xfbase)
{ dim nre = mclxNrofEntries(mxbase)
; mcxLog(MCX_LOG_APP, me, "base has %lu entries", (ulong) nre)
; mclxaWrite(mxbase, xfbase, MCLXIO_VALUE_GETENV, EXIT_ON_FAIL)
; mcxIOclose(xfbase)
; }
if (subcluster_g || dispatch_g)
iaf = iaf ? 1/iaf : 1.414
; while
( (!dispatch_g && (!N || n_ite < N))
|| (dispatch_g && a < argc)
)
{ mclx* mx_coarse = NULL, *clnext = NULL
; dim dist_new_prev = 0, dist_prev_new = 0
; mclx* clnew = NULL
; mcxbool faith = FALSE
; double inflation = -1.0
; if (subcluster_g)
mx_coarse
= subclusterx_g
? mclxBlockPartition(mxbase, clprev, 50)
: mclxBlockUnion(mxbase, clprev)
/* have to copy mxbase as mx_coarse is freed.
* Even if it were not freed, it is probably transformed.
*/
; else if (dispatch_g)
mx_coarse = mclxCopy(mxbase)
; else
{ mx_coarse = get_coarse(mxbase, clprev, add_transpose)
; if (n_ite == 1)
{ mclx* cc = clmUGraphComponents(mx_coarse, NULL) /* fixme; mx_coarse garantueed UD ? */
; n_components = N_COLS(cc)
; mclxFree(&cc)
; }
}
if (xfcoarse)
write_coarse(xfcoarse, mx_coarse)
; get_interface
( &mlp
, NULL
, shared->str
, a < argc ? argv[a] : NULL
, mx_coarse
, ALG_CACHE_START
, EXIT_ON_FAIL
)
; inflation = mlp->mpp->mainInflation
; BIT_OFF(mlp->modes, ALG_DO_SHOW_PID | ALG_DO_SHOW_JURY)
; if ((status = mclAlgorithm(mlp)) == STATUS_FAIL)
{ mcxErr(me, "failed")
; mcxExit(1)
; }
cl_coarse = mclAlgParamRelease(mlp, mlp->cl_result)
; if (xfcoarse)
mclxaWrite(cl_coarse, xfcoarse, MCLXIO_VALUE_NONE, RETURN_ON_FAIL)
; if (dispatch_g || subcluster_g)
clnext = cl_coarse
; else
clnext = mclxCompose(clprev, cl_coarse, 0)
, clnext = control_test(clnext, clctrl)
, mclxFree(&cl_coarse)
; clmSJDistance
(clprev, clnext, NULL, NULL, &dist_prev_new, &dist_new_prev)
; if (dist_prev_new + dist_new_prev)
{ write_clustering
(clnext, clprev, xfcone, xfstack, plexprefix, multiplex_idx++, mlp)
; clnew = clnext
; if (subcluster_g || dispatch_g)
mclxCatPush(&stck_g, clnext, NULL, NULL, mclxCBdomStack, NULL, "dummy-mclcm", n_cls++)
; else
mclxFree(&clprev)
; clprev = clnew
; }
else if
( N_COLS(clnext) > n_components
&& inflation * iaf > 1.2
&& inflation * iaf < 10
)
{ mclxFree(&clnext)
; inflation *= iaf
; mcxTingPrintAfter(shared, " -I %.2f", inflation)
; mcxLog(MCX_LOG_APP, me, "setting inflation to %.2f", inflation)
; faith = TRUE
; }
/* i.e. vanilla mode, contraction */
else if (!subcluster_g && !dispatch_g)
{ mclx* cc
; mclxFree(&clnext)
; mclxAddTranspose(mx_coarse, 1.0)
; cc = clmUGraphComponents(mx_coarse, NULL)
; if (N_COLS(cc) < N_COLS(clprev))
{ mclx* ccback = mclxCompose(clprev, cc, 0)
; write_clustering
(ccback, clprev, xfcone, xfstack, plexprefix, multiplex_idx++, NULL)
; mclxFree(&clprev)
; clprev = ccback
; mcxTell(me, "connected components added as root clustering")
; }
if (root && N_COLS(cc) > 1)
{ mclx* root = mclxCartesian
( mclvCanonical(NULL, 1, 0)
, mclvCopy(NULL, mxbase->dom_cols)
, 1.0
)
; write_clustering
(root, clprev, xfcone, xfstack, plexprefix, multiplex_idx++, NULL)
; mclxFree(&clprev)
; mcxTell(me, "universe added as root clustering")
; clprev = root
; clnew = NULL
; }
mclxFree(&cc)
; }
else if (subcluster_g || dispatch_g)
mclxFree(&clnext)
; mclAlgParamFree(&mlp, TRUE) /* frees mx_coarse */
; if (!clnew && !faith)
{ same = TRUE
; break
; }
a++
; if (dispatch_g && a == argc)
break
; n_ite++
; }
if (same)
mcxLog(MCX_LOG_MODULE, me, "no further contraction: halting")
; if (dispatch_g)
integrate_results(&stck_g)
; else if (subcluster_g)
mclxCatReverse(&stck_g)
; if (dispatch_g || subcluster_g)
{ dim j
; if (xfstack)
mclxCatWrite(xfstack, &stck_g, MCLXIO_VALUE_NONE, RETURN_ON_FAIL)
; if (xfcone && ! mclxCatConify(&stck_g))
mclxCatWrite(xfcone, &stck_g, MCLXIO_VALUE_NONE, RETURN_ON_FAIL)
; for (j=0;j<stck_g.n_level;j++)
{ mclxAnnot* an = stck_g.level+j
; mclxFree(&an->mx)
; }
mcxFree(stck_g.level)
; }
mcxIOfree(&xfcoarse)
; mcxIOfree(&xfbase)
; mcxIOfree(&xfcone)
; mcxIOfree(&xfstack)
; mcxTingFree(&shared)
; if (!dispatch_g && !subcluster_g) /* fixme fixme fixme */
mclxFree(&clprev)
; mclxFree(&mxbase)
; mclvFree(&start_col_sums_g)
; mcxTingFree(&cline)
; helpful_reminder()
; return STATUS_OK
; }
static mcxstatus mateMain
( int argc_unused cpl__unused
, const char* argv[]
)
{ mcxIO* xfx, *xfy
; mclx* mx, *my, *meet, *teem, *myt
; dim x, y
; mcxIOopen(xfout, EXIT_ON_FAIL)
; xfx = mcxIOnew(argv[0], "r")
; mx = mclxRead(xfx, EXIT_ON_FAIL)
; mcxIOclose(xfx)
; xfy = mcxIOnew(argv[1], "r")
; my = mclxRead(xfy, EXIT_ON_FAIL)
; myt = mclxTranspose(my)
; if (!MCLD_EQUAL(mx->dom_rows, my->dom_rows))
mcxDie(1, me, "domains are not equal")
; meet= mclxCompose(myt, mx, 0, 0) /* fixme thread interface */
; teem= mclxTranspose(meet)
; if (legend)
fprintf
( xfout->fp
, "%-10s %6s %6s %6s %6s %6s %6s %6s\n"
, "overlap"
, "x-idx"
, "y-idx"
, "meet"
, "xdiff"
, "ydiff"
, "x-size"
, "y-size"
)
; for (x=0;x<N_COLS(meet);x++)
{ mclv* xvec = meet->cols+x
; long X = xvec->vid
; long xsize = mx->cols[x].n_ivps
; if (one2many && xvec->n_ivps < 2)
continue
; for (y=0;y<N_COLS(teem);y++)
{ mclv* yvec = teem->cols+y
; long Y = yvec->vid
; long ysize = my->cols[y].n_ivps
; double twinfac
; long meetsize
; mclp* ivp = mclvGetIvp(yvec, X, NULL)
; if (!ivp)
continue
/*
* meet size, left diff, right diff, right size.
*/
; meetsize = ivp->val
; if (!xsize && !ysize) /* paranoia */
continue
; twinfac = 2 * meetsize / ( (double) (xsize + ysize) )
; if (xfout)
fprintf
( xfout->fp
, "%-10.3f %6ld %6ld %6ld %6ld %6ld %6ld %6ld\n"
, twinfac
, X
, Y
, meetsize
, xsize - meetsize
, ysize - meetsize
, xsize
, ysize
)
; }
}
return STATUS_OK
; }
int main
( int argc
, const char* argv[]
)
{ mcxIO* xf_tab = NULL
; mcxIO* xf_tabr = NULL
; mcxIO* xf_tabc = NULL
; mcxIO* xf_restrict_tab = NULL
; mcxIO* xf_restrict_tabr = NULL
; mcxIO* xf_restrict_tabc = NULL
; mcxIO* xf_mx = mcxIOnew("-", "r")
; mcxIO* xfout = NULL
; const char* fndump = "-"
; mclTab* tabr = NULL
; mclTab* tabc = NULL
; mclTab* restrict_tabr = NULL
; mclTab* restrict_tabc = NULL
; mcxbool transpose = FALSE
; mcxbool lazy_tab = FALSE
; mcxbool write_tabc = FALSE
; mcxbool write_tabr = FALSE
; mcxbool cat = FALSE
; mcxbool tree = FALSE
; mcxbool skel = FALSE
; mcxbool newick = FALSE
; mcxbits newick_bits = 0
; mcxbits cat_bits = 0
; dim catmax = 1
; dim n_max = 0
; dim table_nlines = 0
; dim table_nfields = 0
; int split_idx = 1
; int split_inc = 1
; const char* split_stem = NULL
; const char* sort_mode = NULL
; mcxTing* line = mcxTingEmpty(NULL, 10)
; mcxbits modes = MCLX_DUMP_VALUES
; mcxbits mode_dump = MCLX_DUMP_PAIRS
; mcxbits mode_part = 0
; mcxbits mode_loop = MCLX_DUMP_LOOP_ASIS
; mcxbits mode_matrix = 0
; int digits = MCLXIO_VALUE_GETENV
; mcxOption* opts, *opt
; mcxstatus parseStatus = STATUS_OK
; mcxLogLevel =
MCX_LOG_AGGR | MCX_LOG_MODULE | MCX_LOG_IO | MCX_LOG_GAUGE | MCX_LOG_WARN
; mclxIOsetQMode("MCLXIOVERBOSITY", MCL_APP_VB_YES)
; mclx_app_init(stderr)
; mcxOptAnchorSortById(options, sizeof(options)/sizeof(mcxOptAnchor) -1)
; opts = mcxOptParse(options, (char**) argv, argc, 1, 0, &parseStatus)
; if (!opts)
exit(0)
; for (opt=opts;opt->anch;opt++)
{ mcxOptAnchor* anch = opt->anch
; switch(anch->id)
{ case MY_OPT_HELP
: case MY_OPT_APROPOS
: mcxOptApropos(stdout, me, syntax, 0, 0, options)
; return 0
;
case MY_OPT_VERSION
: app_report_version(me)
; return 0
;
case MY_OPT_TAB
: xf_tab = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_TABC
: xf_tabc = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_TABR
: xf_tabr = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_OUTPUT
: fndump = opt->val
; break
;
case MY_OPT_SEP_LEAD
: sep_lead_g = opt->val
; break
;
case MY_OPT_SEP_FIELD
: sep_row_g = opt->val
; break
;
case MY_OPT_SEP_CAT
: sep_cat_g = opt->val
; break
;
case MY_OPT_SEP_VAL
: sep_val_g = opt->val
; break
;
case MY_OPT_PREFIXC
: prefixc_g = opt->val
; break
;
case MY_OPT_RESTRICT_TAB
: xf_restrict_tab = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_RESTRICT_TABC
: xf_restrict_tabc = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_RESTRICT_TABR
: xf_restrict_tabr = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_LAZY_TAB
: lazy_tab = TRUE
; break
;
case MY_OPT_NO_VALUES
: BIT_OFF(modes, MCLX_DUMP_VALUES)
; break
;
case MY_OPT_DUMP_RLINES
: mode_dump = MCLX_DUMP_LINES
; BIT_ON(modes, MCLX_DUMP_NOLEAD)
; break
;
case MY_OPT_DUMP_VLINES
: mode_dump = MCLX_DUMP_LINES
; BIT_ON(modes, MCLX_DUMP_LEAD_VALUE)
; break
;
case MY_OPT_DUMP_LINES
: mode_dump = MCLX_DUMP_LINES
; break
;
case MY_OPT_OMIT_EMPTY
: BIT_ON(modes, MCLX_DUMP_OMIT_EMPTY)
; break
;
case MY_OPT_SORT
: sort_mode = opt->val
; break
;
case MY_OPT_NO_LOOPS
: mode_loop = MCLX_DUMP_LOOP_NONE
; break
;
case MY_OPT_CAT_LIMIT
: n_max = atoi(opt->val)
; break
;
case MY_OPT_SPLIT_STEM
: split_stem = opt->val
; sep_cat_g = NULL
; break
;
case MY_OPT_FORCE_LOOPS
: mode_loop = MCLX_DUMP_LOOP_FORCE
; break
;
case MY_OPT_SKEL
: skel = TRUE
; break
;
case MY_OPT_WRITE_TABC
: write_tabc = TRUE
; break
;
case MY_OPT_DIGITS
: digits = strtol(opt->val, NULL, 10)
; break
;
case MY_OPT_WRITE_TABR
: write_tabr = TRUE
; break
;
case MY_OPT_DUMP_RDOM
: transpose = TRUE
; skel = TRUE
; mode_dump = MCLX_DUMP_LINES
; break
;
case MY_OPT_DUMP_CDOM
: skel = TRUE
; mode_dump = MCLX_DUMP_LINES
; break
;
case MY_OPT_IMX
: mcxIOnewName(xf_mx, opt->val)
; break
;
case MY_OPT_ICL
: mcxIOnewName(xf_mx, opt->val)
; mode_dump = MCLX_DUMP_LINES
; BIT_ON(modes, MCLX_DUMP_NOLEAD)
; BIT_OFF(modes, MCLX_DUMP_VALUES)
; break
;
case MY_OPT_TREECAT
: mcxIOnewName(xf_mx, opt->val)
; tree = TRUE
; cat_bits |= MCLX_PRODUCE_DOMSTACK
; break
;
case MY_OPT_CAT
: mcxIOnewName(xf_mx, opt->val)
; cat = TRUE
; break
;
case MY_OPT_DUMP_MATRIX
: mode_matrix |= MCLX_DUMP_MATRIX
; break
;
case MY_OPT_TRANSPOSE
: transpose = TRUE
; break
;
case MY_OPT_DUMP_UPPER
: mode_part = MCLX_DUMP_PART_UPPER
; break
;
case MY_OPT_DUMP_UPPERI
: mode_part = MCLX_DUMP_PART_UPPERI
; break
;
case MY_OPT_DUMP_LOWER
: mode_part = MCLX_DUMP_PART_LOWER
; break
;
case MY_OPT_DUMP_LOWERI
: mode_part = MCLX_DUMP_PART_LOWERI
; break
;
case MY_OPT_DUMP_NOLEAD
: BIT_ON(modes, MCLX_DUMP_NOLEAD)
; break
;
case MY_OPT_NEWICK_MODE
: if (strchr(opt->val, 'N'))
newick_bits |= (MCLX_NEWICK_NONL | MCLX_NEWICK_NOINDENT)
; if (strchr(opt->val, 'I'))
newick_bits |= MCLX_NEWICK_NOINDENT
; if (strchr(opt->val, 'B'))
newick_bits |= MCLX_NEWICK_NONUM
; if (strchr(opt->val, 'S'))
newick_bits |= MCLX_NEWICK_NOPTHS
; newick = TRUE
; break
;
case MY_OPT_DUMP_NEWICK
: newick = TRUE
; break
;
case MY_OPT_DUMP_TABLE
: mode_dump = MCLX_DUMP_TABLE
; break
;
case MY_OPT_TABLE_NFIELDS
: table_nfields = atoi(opt->val)
; break
;
case MY_OPT_TABLE_NLINES
: table_nlines = atoi(opt->val)
; break
;
case MY_OPT_DUMP_PAIRS
: mode_dump = MCLX_DUMP_PAIRS
; break
; }
}
; if (skel)
cat_bits |= MCLX_READ_SKELETON
; modes |= mode_loop | mode_dump | mode_part | mode_matrix
; xfout = mcxIOnew(fndump, "w")
; mcxIOopen(xfout, EXIT_ON_FAIL)
; mcxIOopen(xf_mx, EXIT_ON_FAIL)
; if (cat || tree)
catmax = n_max ? n_max : 0
; if ((write_tabc || write_tabr) && !xf_tab)
mcxDie(1, me, "need a single tab file (-tab option) with --write-tabc or --write-tabr")
; if (xf_tab && mcxIOopen(xf_tab, RETURN_ON_FAIL))
mcxDie(1, me, "no tab")
; else
{ if (xf_tabr && mcxIOopen(xf_tabr, RETURN_ON_FAIL))
mcxDie(1, me, "no tabr")
; if (xf_tabc && mcxIOopen(xf_tabc, RETURN_ON_FAIL))
mcxDie(1, me, "no tabc")
; }
{ if (xf_restrict_tab && mcxIOopen(xf_restrict_tab, RETURN_ON_FAIL))
mcxDie(1, me, "no restriction tab")
; else
{ if (xf_restrict_tabr && mcxIOopen(xf_restrict_tabr, RETURN_ON_FAIL))
mcxDie(1, me, "no restriction tabr")
; if (xf_restrict_tabc && mcxIOopen(xf_restrict_tabc, RETURN_ON_FAIL))
mcxDie(1, me, "no restriction tabc")
; }
/* fixme: below is pretty boilerplate, happens in other places as well */
if (xf_restrict_tab)
{ if (!(restrict_tabr = mclTabRead (xf_restrict_tab, NULL, RETURN_ON_FAIL)))
mcxDie(1, me, "error reading restriction tab")
; restrict_tabc = restrict_tabr
; mcxIOclose(xf_restrict_tab)
; }
else
{ if (xf_restrict_tabr)
{ if (!(restrict_tabr = mclTabRead(xf_restrict_tabr, NULL, RETURN_ON_FAIL)))
mcxDie(1, me, "error reading restriction tabr")
; mcxIOclose(xf_restrict_tabr)
; }
if (xf_restrict_tabc)
{ if (!(restrict_tabc = mclTabRead(xf_restrict_tabc, NULL, RETURN_ON_FAIL)))
mcxDie(1, me, "error reading restriction tabc")
; mcxIOclose(xf_restrict_tabc)
; }
}
}
/* fixme: restructure code to include bit below */
if (write_tabc || write_tabr)
{ mclv* dom_cols = mclvInit(NULL)
; mclv* dom_rows = mclvInit(NULL)
; mclv* dom = write_tabc ? dom_cols : dom_rows
; if (!(tabc = mclTabRead(xf_tab, NULL, RETURN_ON_FAIL)))
mcxDie(1, me, "error reading tab file")
; if (mclxReadDomains(xf_mx, dom_cols, dom_rows))
mcxDie(1, me, "error reading matrix file")
; mcxIOclose(xf_mx)
/* fixme check status */
; mclTabWrite(tabc, xfout, dom, RETURN_ON_FAIL)
; mcxIOclose(xfout)
; return 0
; }
if (newick)
{ mcxTing* thetree
; mclxCat cat
; if (xf_tab && !(tabr = mclTabRead(xf_tab, NULL, RETURN_ON_FAIL)))
mcxDie(1, me, "error reading tab file")
; mclxCatInit(&cat)
; if
( mclxCatRead
( xf_mx
, &cat
, 0
, NULL
, tabr ? tabr->domain : NULL
, MCLX_CATREAD_CLUSTERTREE | MCLX_ENSURE_ROOT
)
)
mcxDie(1, me, "failure reading file")
; thetree = mclxCatNewick(&cat, tabr, newick_bits)
; fwrite(thetree->str, 1, thetree->len, xfout->fp)
; fputc('\n', xfout->fp)
; mcxIOclose(xfout)
; return 0
; }
while (1)
{ mclxIOdumper dumper
; mclxCat cat
; dim i
; if (xf_tab && !lazy_tab)
cat_bits |= MCLX_REQUIRE_GRAPH
; mclxCatInit(&cat)
; if (mclxCatRead(xf_mx, &cat, catmax, NULL, NULL, cat_bits))
break
; for (i=0;i<cat.n_level;i++)
{ mclx* mx = cat.level[i].mx
; if (restrict_tabr || restrict_tabc)
{ mclx* sub
; sub
= mclxSub
( mx
, restrict_tabc
? restrict_tabc->domain
: mx->dom_cols
, restrict_tabr
? restrict_tabr->domain
: mx->dom_rows
)
; mx = sub
; }
/* noteme fixme dangersign mx now may violate some 'cat' invariant */
if (sort_mode)
{ if (!strcmp(sort_mode, "size-ascending"))
mclxColumnsRealign(mx, mclvSizeCmp)
; else if (!strcmp(sort_mode, "size-descending"))
mclxColumnsRealign(mx, mclvSizeRevCmp)
; else
mcxErr(me, "unknown sort mode <%s>", sort_mode)
; if (catmax != 1)
mcxErr(me, "-sort option and cat mode may fail or corrupt")
; }
if (xf_tab && !tabr)
{ if (!( tabr = mclTabRead
(xf_tab, lazy_tab ? NULL : mx->dom_rows, RETURN_ON_FAIL)
) )
mcxDie(1, me, "consider using --lazy-tab option")
; tabc = tabr
; mcxIOclose(xf_tab)
; }
else
{ if (!tabr && xf_tabr)
{ if (!(tabr = mclTabRead
(xf_tabr, lazy_tab ? NULL : mx->dom_rows, RETURN_ON_FAIL)
) )
mcxDie(1, me, "consider using --lazy-tab option")
; mcxIOclose(xf_tabr)
; }
if (!tabc && xf_tabc)
{ if (!( tabc = mclTabRead
(xf_tabc, lazy_tab ? NULL : mx->dom_cols, RETURN_ON_FAIL)
) )
mcxDie(1, me, "consider using --lazy-tab option")
; mcxIOclose(xf_tabc)
; }
}
; if (transpose)
{ mclx* tp = mclxTranspose(mx)
; mclxFree(&mx)
; mx = tp
; if (tabc || tabr)
{ mclTab* tabt = tabc
; tabc = tabr
; tabr = tabt
; }
}
if (mode_dump == MCLX_DUMP_TABLE)
BIT_ON(modes, MCLX_DUMP_TABLE_HEADER)
; mclxIOdumpSet(&dumper, modes, sep_lead_g, sep_row_g, sep_val_g)
; dumper.table_nlines = table_nlines
; dumper.table_nfields = table_nfields
; dumper.prefixc = prefixc_g
; if (split_stem)
{ mcxTing* ting = mcxTingPrint(NULL, "%s.%03d", split_stem, split_idx)
; mcxIOclose(xfout)
; mcxIOrenew(xfout, ting->str, "w")
; split_idx += split_inc
; }
if
( mclxIOdump
( mx
, xfout
, &dumper
, tabc
, tabr
, digits
, RETURN_ON_FAIL
) )
mcxDie(1, me, "something suboptimal")
; mclxFree(&mx)
; if (sep_cat_g && i+1 < cat.n_level)
fprintf(xfout->fp, "%s\n", sep_cat_g)
; }
break
; }
mcxIOfree(&xf_mx)
; mcxIOfree(&xfout)
; mcxIOfree(&xf_tab)
; mcxIOfree(&xf_tabr)
; mcxIOfree(&xf_tabc)
; mcxTingFree(&line)
; return 0
; }
int main
( int argc
, const char* argv[]
)
{ mcxIO* xfdagreduce = NULL, *xfattr = NULL, *xfdiff = NULL
; double child_diff_lq = 0.2
; double parent_diff_gq = 0.4
; mcxIO* xfimx = mcxIOnew("-", "r"), *xfdag = NULL, *xftab = NULL
; mclTab* tab = NULL
; int q = -1
; mclx* mx
; unsigned char test_mode = 0
; mcxstatus parseStatus = STATUS_OK
; mcxOption* opts, *opt
; mcxOptAnchorSortById(options, sizeof(options)/sizeof(mcxOptAnchor) -1)
; if
(!(opts = mcxOptParse(options, (char**) argv, argc, 1, 0, &parseStatus)))
exit(0)
; mcxLogLevel =
MCX_LOG_AGGR | MCX_LOG_MODULE | MCX_LOG_IO | MCX_LOG_GAUGE | MCX_LOG_WARN
; mclxIOsetQMode("MCLXIOVERBOSITY", MCL_APP_VB_YES)
; mclx_app_init(stderr)
; for (opt=opts;opt->anch;opt++)
{ mcxOptAnchor* anch = opt->anch
; switch(anch->id)
{ case MY_OPT_HELP
: mcxOptApropos(stdout, me, syntax, 0, 0, options)
; return 0
;
case MY_OPT_VERSION
: app_report_version(me)
; return 0
;
case MY_OPT_TEST_CYCLE
: test_mode = 'c'
; break
;
case MY_OPT_TEST_CROSS
: test_mode = 'x'
; break
;
case MY_OPT_DAG_ATTR
: xfattr = mcxIOnew(opt->val, "w")
; mcxIOopen(xfattr, EXIT_ON_FAIL)
; break
;
case MY_OPT_DAG_DIFF
: xfdiff = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_DAG_REDUCE
: xfdagreduce = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_CHILD_DIFF_LQ
: child_diff_lq = atof(opt->val)
; break
;
case MY_OPT_PARENT_DIFF_GQ
: parent_diff_gq = atof(opt->val)
; break
;
case MY_OPT_QUERY
: q = atoi(opt->val)
; break
;
case MY_OPT_TAB
: xftab = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_IMX
: mcxIOnewName(xfimx, opt->val)
; break
;
}
}
; if (xfimx)
mx = mclxRead(xfimx, EXIT_ON_FAIL)
; else
mcxDie(1, me, "need -imx")
; if (xftab)
tab = mclTabRead(xftab, mx->dom_cols, EXIT_ON_FAIL)
; if (test_mode == 'c')
test_for_cycles(mx)
; else if (test_mode == 'x')
test_cross_ratio(mx)
; else if (xfattr)
get_attr(mx, tab, xfattr)
; else if (xfdagreduce)
{ mclxComposeHelper* ch = mclxComposePrepare(mx, mx)
; dim i
; for (i=0;i<N_COLS(mx);i++)
{ mclv* in = mx->cols+i
; mclv* out = mclxVectorCompose(mx, in, NULL, ch)
; mcldMinus(in, out, in)
; mclvFree(&out)
; }
mclxWrite(mx, xfdagreduce, MCLXIO_VALUE_GETENV, EXIT_ON_FAIL)
; mclxComposeRelease(&ch)
; }
else if (xfdiff)
dag_diff_select(mx, tab, xfdiff, child_diff_lq, parent_diff_gq)
; mclxFree(&mx)
; mcxIOfree(&xfimx)
; mcxIOfree(&xfdag)
; mcxIOfree(&xfattr)
; mcxIOfree(&xfdagreduce)
; return 0
; }
static mcxstatus collectMain
( int argc
, const char* argv[]
)
{ aggr* collect = NULL
; int a
; dim i, collect_n = 0
; mclTab* tab = NULL
; double avg = 0.0
; mclx* aggr = NULL, *mx = NULL
/* mcxHash* map = NULL */
; mcxIO* xfout = mcxIOnew(out_g, "w")
; mcxIOopen(xfout, EXIT_ON_FAIL)
; if
( transform_spec
&& (!(transform = mclgTFparse(NULL, transform_spec)))
)
mcxDie(1, me, "input -tf spec does not parse")
; if (xftab_g)
tab = mclTabRead(xftab_g, NULL, EXIT_ON_FAIL)
/* map not used; perhaps someday we want to map labels to indexes?
* in that case, we could also simply reverse the tab when reading ..
, map = mclTabHash(tab)
*/
; if (!collect_g)
mcxDie(1, me, "require one of --paste, --add-column, --add-matrix")
; if (argc)
{ if (collect_g == 'm')
{ mcxIO* xf = mcxIOnew(argv[0], "r")
; mcxIOopen(xf, EXIT_ON_FAIL)
; aggr = mclxRead(xf, EXIT_ON_FAIL)
; mcxIOfree(&xf)
; }
else
collect_n = do_a_file(&collect, argv[0], 0)
; }
if (tab && collect_n != N_TAB(tab) + (header_g ? 1 : 0))
mcxErr
( me
, "tab has differing size (%lu vs %lu), continuing anyway"
, (ulong) N_TAB(tab)
, (ulong) (collect_n ? collect_n -1 : 0)
)
; for (a=1;a<argc;a++)
{ if (collect_g == 'm')
{ mcxIO* xf = mcxIOnew(argv[a], "r")
; mcxIOopen(xf, EXIT_ON_FAIL)
; mx = mclxRead(xf, EXIT_ON_FAIL)
; mclxAugment(aggr, mx, fltop_g)
; mcxIOfree(&xf)
; mclxFree(&mx)
; }
else
do_a_file(&collect, argv[a], collect_n)
; }
if (collect_g == 'm')
{ if (transform)
mclgTFexec(aggr, transform)
; if (mcx_wb_g)
mclxbWrite(aggr, xfout, EXIT_ON_FAIL)
; else
mclxWrite(aggr, xfout, MCLXIO_VALUE_GETENV, EXIT_ON_FAIL)
; mcxIOclose(xfout)
; exit(0)
; }
/* fimxe: dispatch on binary_g */
for (i=0;i<collect_n;i++)
{ const char* lb = collect[i].label
; if (!i && collect[i].columns && collect_g != 'p')
{ fprintf(xfout->fp, "%s\t%s\n", lb, collect[i].columns->str)
; continue
; }
if (tab && (!header_g || i > 0))
{ unsigned u = atoi(lb)
; lb = mclTabGet(tab, u, NULL)
; if (TAB_IS_NA(tab, lb))
mcxDie(1, me, "no label found for index %ld - abort", (long) u)
; }
if (summary_g)
avg += collect[i].val
; else
{ if (collect_g == 'p')
fprintf(xfout->fp, "%s%s\n", lb, collect[i].columns->str)
; else
fprintf(xfout->fp, "%s\t%.8g\n", lb, collect[i].val)
; }
}
if (summary_g && collect_n)
{ dim middle1 = (collect_n-1)/2, middle2 = collect_n/2
; qsort(collect, collect_n, sizeof collect[0], aggr_cmp_val)
; avg /= collect_n
; fprintf /* --summary option is a bit rubbish interface-wise */
( xfout->fp
, "%g %g %g %g\n"
, collect[0].val
, (collect[middle1].val + collect[middle2].val) / 2
, collect[collect_n-1].val
, avg
)
; }
return STATUS_OK
; }
static dim do_a_file
( aggr** collectpp
, const char* fname
, dim collect_n
)
{ mcxIO* xf = mcxIOnew(fname, "r")
; mcxTing* buf = mcxTingEmpty(NULL, 100)
; mcxTing* lbl = mcxTingEmpty(NULL, 100)
; mcxstatus status = STATUS_OK
; aggr* collect = *collectpp
; dim ct = 0, collect_alloc = 0
; if (!collect_n)
collect_alloc = 100
, collect = mcxNAlloc(collect_alloc, sizeof collect[0], NULL, EXIT_ON_FAIL)
; mcxIOopen(xf, EXIT_ON_FAIL)
; while (STATUS_OK == (status = mcxIOreadLine(xf, buf, MCX_READLINE_CHOMP)))
{ double val
; const char* tabchar = NULL
; mcxbool get_header = collect_g != 'p' && !ct ? TRUE : FALSE
; mcxTingEnsure(lbl, buf->len)
/* if header_g && !ct && !paste create/check label */
/* body of this while loop does too many things, refactor */
; if (collect_g == 'p' || get_header)
{ if (!(tabchar = strchr(buf->str, '\t')))
mcxDie(1, me, "paste error at line %d file %s (no tab)", (int) xf->lc, fname)
; mcxTingNWrite(lbl, buf->str, tabchar - buf->str)
; }
else
{ if (2 != sscanf(buf->str, "%s%lg", lbl->str, &val))
mcxDie(1, me, "parse error at line %d file %s", (int) xf->lc, fname)
; lbl->len = strlen(lbl->str)
; }
if (!collect_n)
{ if (ct >= collect_alloc)
{ dim collect_realloc = collect_alloc * 1.44
; collect = mcxNRealloc(collect, collect_realloc, collect_alloc, sizeof collect[0], NULL, EXIT_ON_FAIL)
; collect_alloc = collect_realloc
; }
collect[ct].label = mcxTingStr(lbl)
; collect[ct].val = collect_g == 'p' || get_header ? 0.0 : val
; collect[ct].columns
= collect_g == 'p' || get_header
? mcxTingNew(tabchar + (get_header ? 1 : 0))
: NULL
; }
else
{ if (ct >= collect_n)
mcxDie(1, me, "additional lines in file %s", fname)
; if (strcmp(collect[ct].label, lbl->str))
mcxDie
( 1
, me
, "label conflict %s/%s at line %d in file %s"
, collect[ct].label
, lbl->str
, (int) xf->lc, fname
)
; if (get_header) /* only need to check identity */
{ if (strcmp(tabchar+1, collect[ct].columns->str))
mcxDie(1, me, "different columns <%s> and <%s>", collect[ct].columns->str, tabchar+1)
; }
else if (collect_g == 'p') /* tack it on */
mcxTingNAppend(collect[ct].columns, tabchar, buf->len - lbl->len)
; else
collect[ct].val += val
; }
ct++
; }
if (collect_n)
{ if (ct != collect_n)
mcxDie(1, me, "not enough lines in file %s", fname)
; }
else
{ if (!ct)
mcxDie(1, me, "empty file(s)")
; *collectpp = collect
; }
mcxIOfree(&xf)
; return ct
; }
int main
( int argc
, const char* argv[]
)
{ mcxIO* xfmx = mcxIOnew("-", "r"), *xfout = mcxIOnew("-", "w")
; mclx* mx = NULL
; mclv* mx_diag = NULL
; mcxstatus parseStatus = STATUS_OK
; mcxOption* opts, *opt
; dim N_edge = 0
; dim* offsets
; dim template_n_nodes = 0
; mcxbool plus = FALSE
; double e_min = 1.0
; double e_max = 0.0
; double skew = 0.0
; double radius = 0.0
; double n_sdev = 0.5
; double n_range = 2.0
; double g_radius = 0.0
; double g_mean = 0.0
; double g_sdev = 0.0
; double g_min = 1.0
; double g_max = 0.0
; mcxbool do_gaussian = FALSE
; dim i = 0
; dim N_remove = 0
; dim N_add = 0
; dim N_shuffle = 0
; unsigned long random_ignore = 0
; srandom(mcxSeed(2308947))
; mcxOptAnchorSortById(options, sizeof(options)/sizeof(mcxOptAnchor) -1)
; if
(!(opts = mcxOptParse(options, (char**) argv, argc, 1, 0, &parseStatus)))
exit(0)
; mcxLogLevel =
MCX_LOG_AGGR | MCX_LOG_MODULE | MCX_LOG_IO | MCX_LOG_GAUGE | MCX_LOG_WARN
; mclxIOsetQMode("MCLXIOVERBOSITY", MCL_APP_VB_YES)
; mclx_app_init(stderr)
; for (opt=opts;opt->anch;opt++)
{ mcxOptAnchor* anch = opt->anch
; switch(anch->id)
{ case MY_OPT_HELP
: case MY_OPT_APROPOS
: mcxOptApropos(stdout, me, syntax, 20, MCX_OPT_DISPLAY_SKIP, options)
; return 0
;
case MY_OPT_VERSION
: app_report_version(me)
; return 0
;
case MY_OPT_SKEW
: skew = atof(opt->val)
; break
;
case MY_OPT_GEN
: template_n_nodes = atoi(opt->val)
; break
;
case MY_OPT_IMX
: mcxIOrenew(xfmx, opt->val, NULL)
; break
;
case MY_OPT_PLUS
: case MY_OPT_WB
: plus = TRUE
; break
;
case MY_OPT_OUT
: mcxIOrenew(xfout, opt->val, NULL)
; break
;
case MY_OPT_E_MAX
: if (!strcmp(opt->val, "copy"))
e_max = -DBL_MAX
; else
e_max = atof(opt->val)
; break
;
case MY_OPT_E_MIN
: e_min = atof(opt->val)
; break
;
case MY_OPT_G_MIN
: g_min = atof(opt->val)
; break
;
case MY_OPT_G_MAX
: g_max = atof(opt->val)
; break
;
case MY_OPT_G_SDEV
: g_sdev = atof(opt->val)
; break
;
case MY_OPT_G_MEAN
: g_mean = atof(opt->val)
; do_gaussian = TRUE
; break
;
case MY_OPT_G_RADIUS
: g_radius = atof(opt->val)
; break
;
case MY_OPT_N_RANGE
: n_range = atof(opt->val)
; break
;
case MY_OPT_N_SDEV
: n_sdev = atof(opt->val)
; break
;
case MY_OPT_N_RADIUS
: radius = atof(opt->val)
; break
;
case MY_OPT_SHUFFLE
: N_shuffle = atoi(opt->val)
; break
;
case MY_OPT_ADD
: N_add = atoi(opt->val)
; break
;
case MY_OPT_REMOVE
: N_remove = atoi(opt->val)
; break
; }
}
/* hitting y% in vi tells me the size of this block */
{ if (template_n_nodes)
mx = mclxAllocZero
( mclvCanonical(NULL, template_n_nodes, 1.0)
, mclvCanonical(NULL, template_n_nodes, 1.0)
)
; else
mx = mclxReadx
( xfmx
, EXIT_ON_FAIL
, MCLX_REQUIRE_GRAPH
)
; mx_diag = mclxDiagValues(mx, MCL_VECTOR_COMPLETE)
; if (N_shuffle)
mclxAdjustLoops(mx, mclxLoopCBremove, NULL)
; else
mclxSelectUpper(mx)
/* ^ apparently we always work on single arc representation (docme andsoon) */
; offsets = mcxAlloc(sizeof offsets[0] * N_COLS(mx), EXIT_ON_FAIL)
; N_edge = 0
; for (i=0;i<N_COLS(mx);i++)
{ offsets[i] = N_edge
; N_edge += mx->cols[i].n_ivps
; }
if (N_edge < N_remove)
{ mcxErr
( me
, "removal count %ld exceeds edge count %ld"
, (long) N_remove
, (long) N_edge
)
; N_remove = N_edge
; }
random_ignore = RAND_MAX - (N_edge ? RAND_MAX % N_edge : 0)
; if (RAND_MAX / 2 < N_edge)
mcxDie(1, me, "graph too large!")
; if (N_shuffle)
{ do_the_shuffle(mx, N_shuffle, offsets, N_edge, random_ignore)
; mx_readd_diagonal(mx, mx_diag)
; mclxWrite(mx, xfout, MCLXIO_VALUE_GETENV, RETURN_ON_FAIL)
; exit(0)
; }
; if (N_remove)
{ dim n_remove = do_remove(mx, N_remove, offsets, N_edge, random_ignore)
/* Need to recompute N_edge and random_ignore.
* NOTE we work with *upper* matrix; this counts graph edges.
*/
; N_edge = mclxNrofEntries(mx) - n_remove
; random_ignore = RAND_MAX - (RAND_MAX % N_COLS(mx))
; }
if (g_mean)
{ if (!g_radius)
{ if (g_sdev)
g_radius = 2 * g_sdev
; mcxWarn(me, "set radius to %.5f\n", g_radius)
; }
}
; if (N_add)
N_edge += do_add
( mx
, N_add , N_edge
, do_gaussian ? &g_mean : NULL, g_radius , g_sdev , g_min , g_max
, skew
, e_min , e_max
)
; if (radius)
{ for (i=0;i<N_COLS(mx);i++)
{ mclp* ivp = mx->cols[i].ivps, *ivpmax = ivp + mx->cols[i].n_ivps
;if(DEBUG)fprintf(stderr, "here %d\n", (int) i)
; while (ivp < ivpmax)
{ double val = ivp->val
; double r = mcxNormalCut(n_range * n_sdev, n_sdev)
; double newval = val + radius * (r / (n_range * n_sdev))
; if (e_min < e_max && newval >= e_min && newval <= e_max)
; ivp->val = newval
; ivp++
; }
}
}
mclxUnary(mx, fltxCopy, NULL) /* remove zeroes */
; mclxAddTranspose(mx, 0.0)
; mx_readd_diagonal(mx, mx_diag)
; if (plus)
mclxbWrite(mx, xfout, RETURN_ON_FAIL)
; else
mclxWrite(mx, xfout, MCLXIO_VALUE_GETENV, RETURN_ON_FAIL)
; }
return 0
; }
static mcxstatus qArgHandle
( int optid
, const char* val
)
{ switch(optid)
{ case MY_OPT_IMX
: mcxIOnewName(xfmx_g, val)
; user_imx = TRUE
; break
;
case MY_OPT_ABC
: xfabc_g = mcxIOnew(val, "r")
; break
;
case MY_OPT_TAB
: xftab_g = mcxIOnew(val, "r")
; break
;
case MY_OPT_ICL
: xfcl_g = mcxIOnew(val, "r")
; break
;
case MY_OPT_TRANSFORM
: transform_spec = mcxTingNew(val)
; break
;
case MY_OPT_DIMENSION
: mode_get = 'd'
; break
;
case MY_OPT_TESTMETRIC
: mode_get = 'm'
; break
;
case MY_OPT_TESTCYCLE
: mode_get = 'c'
; break
;
case MY_OPT_TESTCYCLE_N
: n_limit = atoi(val)
; mode_get = 'c'
; break
;
case MY_OPT_INFO
: compute_flags |= COMPUTE_EFF
; break
;
case MY_OPT_CLCF
: compute_flags |= COMPUTE_CLCF
; break
;
case MY_OPT_REDUCE
: rebase_g = FALSE
; break
;
case MY_OPT_THREAD
: n_thread_l = atoi(val)
; break
;
case MY_OPT_NODE
: mode_get = 'n'
; break
;
case MY_OPT_FOUT
: mcxIOnewName(xfout_g, val)
; break
;
case MY_OPT_WEIGHT_SCALE
: weight_scale = atof(val)
; break
;
case MY_OPT_OUTPUT_NOLEGEND
: output_flags ^= OUTPUT_KEY
; break
;
case MY_OPT_OUTPUT_TABLE
: output_flags |= OUTPUT_TABLE
; break
;
case MY_OPT_SCALEFREE
: weefreemen = TRUE
; break
;
case MY_OPT_EDGEWEIGHTS
: mode_get = 'e'
; break
;
case MY_OPT_EDGEWEIGHTS_HIST
: mode_get = 'H'
; if (4 != sscanf(val, "%d,%d,%d,%d", &vary_a, &vary_z, &vary_s, &vary_n))
mcxDie(1, me, "failed to parse argument as integers start,end,step,norm")
; break
;
case MY_OPT_EDGEWEIGHTS_SORTED
: mode_get = 'E'
; break
;
case MY_OPT_DIVIDE
: divide_g = atoi(val)
; break
;
case MY_OPT_VARY_CORRELATION
: vary_a = 4
; vary_z = 20
; vary_s = 1
; vary_n = 20
; weight_scale = 100
; mode_vary = 'c'
; break
;
case MY_OPT_VARY_CEIL
: case MY_OPT_VARY_KNN
: case MY_OPT_VARY_N
: case MY_OPT_VARY_THRESHOLD
: if (4 != sscanf(val, "%d,%d,%d,%d", &vary_a, &vary_z, &vary_s, &vary_n))
mcxDie(1, me, "failed to parse argument as integers start,end,step,norm")
; mode_vary
= optid == MY_OPT_VARY_THRESHOLD
? 't'
: optid == MY_OPT_VARY_KNN
? 'k'
: optid == MY_OPT_VARY_CEIL
? 'n'
: optid == MY_OPT_VARY_N
? 'l'
: 'X'
; break
;
default
: mcxExit(1)
;
; }
if (mode_vary && 1000 * vary_s < vary_z - vary_a)
mcxDie(1, me, "argument leads to more than one thousand steps")
; if (!vary_n)
mcxDie(1, me, "need nonzero scaling factor (last component)")
; return STATUS_OK
; }
static mcxstatus qArgHandle
( int optid
, const char* val
)
{ switch(optid)
{ case MY_OPT_IMX
: mcxIOnewName(xfmx_g, val)
; user_imx = TRUE
; break
;
case MY_OPT_TAB
: xftab_g = mcxIOnew(val, "r")
; break
;
case MY_OPT_ICL
: xfcl_g = mcxIOnew(val, "r")
; break
;
case MY_OPT_TRANSFORM
: transform_spec = mcxTingNew(val)
; break
;
case MY_OPT_DIMENSION
: mode_get = 'd'
; break
;
case MY_OPT_TESTCYCLE
: mode_get = 'c'
; break
;
case MY_OPT_TESTCYCLE_N
: n_limit = atoi(val)
; mode_get = 'c'
; break
;
case MY_OPT_CLCF
: doclcf_g = TRUE
; break
;
case MY_OPT_KNNREDUCE
: knnexact_g = FALSE
; break
;
case MY_OPT_THREAD
: n_thread_g = atoi(val)
; break
;
case MY_OPT_NODE
: mode_get = 'n'
; break
;
case MY_OPT_FOUT
: mcxIOnewName(xfout_g, val)
; break
;
case MY_OPT_WEIGHT_SCALE
: weight_scale = atof(val)
; break
;
case MY_OPT_OUTPUT_TABLE
: output_table = TRUE
; break
;
case MY_OPT_MYTH
: weefreemen = TRUE
; break
;
case MY_OPT_EDGEWEIGHTS
: mode_get = 'e'
; break
;
case MY_OPT_EDGEWEIGHTS_HIST
: mode_get = 'H'
; if (4 != sscanf(val, "%d,%d,%d,%d", &vary_a, &vary_z, &vary_s, &vary_n))
mcxDie(1, me, "failed to parse argument as integers start,end,step,norm")
; break
;
case MY_OPT_EDGEWEIGHTS_SORTED
: mode_get = 'E'
; break
;
case MY_OPT_DIVIDE
: divide_g = atoi(val)
; break
;
case MY_OPT_VARY_CORRELATION
: vary_a = 4
; vary_z = 20
; vary_s = 1
; vary_n = 20
; weight_scale = 100
; mode_vary = 'c'
; break
;
case MY_OPT_VARY_KNN
: if (4 != sscanf(val, "%d,%d,%d,%d", &vary_a, &vary_z, &vary_s, &vary_n))
mcxDie(1, me, "failed to parse argument as integers start,end,step,norm")
; mode_vary = 'k'
; break
;
case MY_OPT_VARY_THRESHOLD
: if (4 != sscanf(val, "%u,%u,%u,%u", &vary_a, &vary_z, &vary_s, &vary_n))
mcxDie(1, me, "failed to parse argument as integers start,end,step,norm")
; mode_vary = 't'
; break
;
default
: mcxExit(1)
;
; }
if (mode_vary && 1000 * vary_s < vary_z - vary_a)
mcxDie(1, me, "argument leads to more than one thousand steps")
; if (!vary_n)
mcxDie(1, me, "need nonzero scaling factor (last component)")
; return STATUS_OK
; }
static mcxstatus meetMain
( int argc
, const char* argv[]
)
{ mcxIO **xfmcs = NULL
; mclMatrix *lft = NULL
; mclMatrix *rgt = NULL
; mclMatrix *dst = NULL
; int a = 0
; int n_mx = 0
; int j
; dim o, m, e
; mclxIOsetQMode("MCLXIOVERBOSITY", MCL_APP_VB_YES)
; mclx_app_init(stderr)
; xfmcs = (mcxIO**) mcxAlloc
( (argc)*sizeof(mcxIO*)
, EXIT_ON_FAIL
)
; mcxIOopen(xfout, EXIT_ON_FAIL)
; for(j=a;j<argc;j++)
{ xfmcs[n_mx] = mcxIOnew(argv[j], "r")
; n_mx++
; }
if (!n_mx)
mcxDie(1, me, "at least one clustering matrix required")
/* Fixme: do a decent initialization with lft = clmTop() *before*
* this loop (removing the need for ugly tmp assignment), but that requires
* we know the correct domain to pass to it. For that, we need to peak into
* the first matrix.
*/
; for (j=0;j<n_mx;j++)
{ mclMatrix* tmp = mclxRead (xfmcs[j], EXIT_ON_FAIL)
; if (clmEnstrict(tmp, &o, &m, &e, ENSTRICT_SPLIT_OVERLAP))
report_partition("clmmeet", tmp, xfmcs[j]->fn, o, m, e)
, mcxExit(1)
; if (!lft)
{ lft = tmp
; continue
; }
else
rgt = tmp
; if (!MCLD_EQUAL(lft->dom_rows, rgt->dom_rows))
mcxDie
( 1
, me
, "domains not equal (files %s/%s)"
, xfmcs[j-1]->fn->str
, xfmcs[j]->fn->str
)
; mcxIOclose(xfmcs[j])
; dst = clmMeet(lft, rgt)
; lft = dst
; mclxFree(&rgt)
; }
mclxColumnsRealign(lft, mclvSizeRevCmp)
; mclxWrite(lft, xfout, MCLXIO_VALUE_NONE, EXIT_ON_FAIL)
; mclxFree(&lft)
; mcxIOfree(&xfout)
; free(xfmcs)
; return STATUS_OK
; }
/* Purpose: read a single x/y combination. The x may be cached
* due to the etc format, where a single line always refers to the same x
* and that x is listed only at the start or line, or omitted with
* the etc-ai and 235-ai formats.
*
* state->x_prev may be used by read_etc in order to obtain the
* current x index.
*/
static mcxstatus read_etc
( mcxIO* xf
, stream_state *iface
, etc_state *state
, double* value
)
{ mcxbits bits = iface->bits
; FILE* stdbug = stdout
; mcxstatus status = STATUS_OK
; mcxTing* ykey = NULL
; mcxTing* xkey = NULL
; const char* printable
; mcxbool label_cbits = bits & (MCLXIO_STREAM_CTAB_STRICT | MCLXIO_STREAM_CTAB_RESTRICT)
; mcxbool label_rbits = bits & (MCLXIO_STREAM_RTAB_STRICT | MCLXIO_STREAM_RTAB_RESTRICT)
; mcxbool label_dbits = bits & (MCLXIO_STREAM_WARN | MCLXIO_STREAM_DEBUG)
; iface->statusx = STATUS_OK
; iface->statusy = STATUS_OK
; iface->x = state->x_prev
; *value = 1.0
;if(DEBUG)fprintf(stdbug, "read_etc initially set x to %d\n", (int) iface->x)
; if (!state->etcbuf)
state->etcbuf = mcxTingEmpty(NULL, 100)
; do
{ int n_char_read = 0
; if (state->etcbuf->len != state->etcbuf_check)
{ mcxErr
( module
, "read_etc sanity check failed %ld %ld"
, (long) state->etcbuf->len
, (long) state->etcbuf_check
)
; status = STATUS_FAIL
; break
; }
/* do we need to read a line ? */
/* -> then set iface->x */
/* fixmefixme: funcify this */
/* iface->x can only be changed in this branch */
/* ************************************************************************** */
if (state->etcbuf_ofs >= state->etcbuf->len)
{ state->etcbuf_ofs = 0
; state->n_y = 0
; if ((status = mcxIOreadLine(xf, state->etcbuf, MCX_READLINE_CHOMP)))
break
; state->etcbuf_check = state->etcbuf->len
; if
( !(printable = mcxStrChrAint(state->etcbuf->str, isspace, -1))
|| (unsigned char) *printable == '#'
)
{ state->etcbuf_ofs = state->etcbuf->len
; iface->statusy = STATUS_IGNORE
; break /* fixme: ^ statusx seems to work as well. cleanify design */
; }
; if (bits & (MCLXIO_STREAM_ETC_AI | MCLXIO_STREAM_235_AI))
{
}
/* In this branch we do not issue handle_label, so we take care of max_seen.
*/
else if (bits & MCLXIO_STREAM_235)
{ if (1 != sscanf(state->etcbuf->str, "%lu%n", &(iface->x), &n_char_read))
{ iface->statusx = STATUS_FAIL
; break
; }
state->etcbuf_ofs += n_char_read
; if (iface->map_c->max_seen+1 < iface->x+1) /* note mixed-sign comparison */
iface->map_c->max_seen = iface->x
; state->x_prev = iface->x
; }
else if (bits & MCLXIO_STREAM_ETC)
{ xkey = mcxTingEmpty(NULL, state->etcbuf->len)
; if (1 != sscanf(state->etcbuf->str, "%s%n", xkey->str, &n_char_read))
break
; state->etcbuf_ofs += n_char_read
; xkey->len = strlen(xkey->str)
; xkey->str[xkey->len] = '\0'
;if(DEBUG3)fprintf(stderr, "max %lu\n", (ulong) iface->map_c->max_seen)
; iface->statusx
= handle_label(&xkey, &(iface->x), iface->map_c, label_cbits | label_dbits, "col")
;if(DEBUG3)fprintf(stderr, "max %lu x %lu\n", (ulong) iface->map_c->max_seen, (ulong) iface->x)
; if (iface->statusx == STATUS_IGNORE || iface->statusx == STATUS_FAIL)
{ /* iface->x = 141414 recentlyadded */
;if(DEBUG3)fprintf(stderr, "max %lu\n", (ulong) iface->map_c->max_seen)
; break
; }
/* ^ Consider what happens when we break here (x label not
* accepted) with map_c->max_seen. Basically x label is
* indepedent of y, so we never need to undo the
* handle_label action.
*/
state->x_prev = iface->x
; }
else
mcxDie(1, module, "strange, really")
; }
/* ************************************************************************** */
if
( !( printable
= mcxStrChrAint(state->etcbuf->str+state->etcbuf_ofs, isspace, -1)
)
|| (uchar) *printable == '#'
)
{ state->etcbuf_ofs = state->etcbuf->len
; /* iface->y = 141414 recentlyadded */
; iface->statusy = STATUS_IGNORE
; break
; }
if (bits & (MCLXIO_STREAM_235_AI | MCLXIO_STREAM_235))
{ if (1 != sscanf(state->etcbuf->str+state->etcbuf_ofs, "%lu%n", &(iface->y), &n_char_read))
{ char* s = state->etcbuf->str+state->etcbuf_ofs
; while(isspace((uchar) s[0]))
s++
; mcxErr
( module
, "unexpected string starting with <%c> on line %lu"
, (int) ((uchar) s[0])
, xf->lc
)
; iface->statusy = STATUS_FAIL
; }
else
{
;if(DEBUG3)fprintf(stdbug, "hit at %d\n", (int) state->etcbuf_ofs);
state->etcbuf_ofs += n_char_read
; if (iface->map_r->max_seen+1 < iface->y+1) /* note mixed-sign comparison */
iface->map_r->max_seen = iface->y
; }
}
else /* ETCANY */
{ ykey = mcxTingEmpty(NULL, state->etcbuf->len)
; if (1 != sscanf(state->etcbuf->str+state->etcbuf_ofs, "%s%n", ykey->str, &n_char_read))
break
; ykey->len = strlen(ykey->str)
; ykey->str[ykey->len] = '\0'
; state->etcbuf_ofs += n_char_read
; iface->statusy
= handle_label(&ykey, &(iface->y), iface->map_r, label_rbits | label_dbits, "row")
; }
/* this won't scale well in terms of organisation if and when
* tabs are allowed with 235 mode, because in that case,
* with 235-ai and restrict-tabr and extend-tabc we will
* need the stuff below duplicated in the 235 branch above.
* what happens here is that we only decide now whether
* the auto-increment is actually happening. It depends
* on there being at least one y that was not rejected.
*/
; if
( (bits & (MCLXIO_STREAM_ETC_AI | MCLXIO_STREAM_235_AI))
&& (iface->statusy == STATUS_OK || iface->statusy == STATUS_NEW)
&& !state->n_y
)
{ iface->x = state->x_prev + 1 /* works first time around */
; iface->map_c->max_seen = state->x_prev + 1
; state->n_y++
; state->x_prev = iface->x
; }
;if(DEBUG2)fprintf(stdbug, "etc handle label we have y %d status %s\n", (int) iface->y, MCXSTATUS(iface->statusy))
; }
while (0)
;if(DEBUG2)fprintf(stdbug, "status %s\n", MCXSTATUS(status))
; do
{ if (status) /* e.g. STATUS_DONE (readline) [or STATUS_IGNORE (#)]*/
break
/* below iface->statusy == STATUS_NEW should be impossible
* given this clause and the code sequence earlier.
*/
; if (iface->statusx == STATUS_FAIL || iface->statusx == STATUS_IGNORE)
{ mcxTingFree(&xkey)
; status = iface->statusx
; break
; }
/* case iface->statusx == STATUS_NEW is *always* honored
*/
; if (iface->statusy == STATUS_FAIL || iface->statusy == STATUS_IGNORE)
{ mcxTingFree(&ykey)
; status = iface->statusy
; break
; }
}
while (0)
; if (status == STATUS_IGNORE || status == STATUS_FAIL)
mcxTingFree(&ykey)
/* fixme, the action in this branch is done in other places too. cleanify design */
; if
( iface->statusx == STATUS_IGNORE
|| !mcxStrChrAint(state->etcbuf->str+state->etcbuf_ofs, isspace, -1)
)
state->etcbuf_ofs = state->etcbuf->len
;if(DEBUG3)fprintf
( stdbug, "read_etc %s return x(%s -> %d stat=%s) y(%s -> %d stat=%s) status %s buf %d %d c_max_seen %lu\n"
, MCXSTATUS(status)
, (xkey ? xkey->str : "-"), (int) iface->x, MCXSTATUS(iface->statusx)
, (ykey ? ykey->str : "-"), (int) iface->y, MCXSTATUS(iface->statusy)
, MCXSTATUS(status), (int) state->etcbuf->len, (int) state->etcbuf_ofs
, (ulong) iface->map_c->max_seen
)
; return status
; }
void pairwise_setops
( mclx* mx1
, mclx* mx2
, mcxbits modes
)
{ dim t, u, n_tst = 0
; mclv* cache = mclvInit(NULL)
; mclv* meet = mclvInit(NULL)
; mclv* join = mclvInit(NULL)
; mclv* diff = mclvInit(NULL)
; mcxbool overwrite = modes & MMM_OVERWRITE
; dim n_zero_meet = 0, n_plus_meet = 0
; mclv* (*fn_meet)(const mclv* lft, const mclv* rgt, mclv* dst) = mcldMeet
; mclv* (*fn_minus)(const mclv* lft, const mclv* rgt, mclv* dst) = mcldMinus1
; if (modes & MMM_MEET2)
fn_meet = mcldMeet2
, fn_minus = mcldMinus
/* the point of overwrite is to have
* a lft == dst or rgt == dst pattern.
*/
; for (t=0;t<N_COLS(mx1);t++)
{ for (u=0;u<N_COLS(mx2);u++)
{ mclv* dst = overwrite ? (modes & MMM_RIGHT ? mx1->cols+u : mx2->cols+t) : diff
; if (overwrite)
mclvCopy(cache, dst) /* cache column, reinstate later */
; if (modes & MMM_BINARY)
mclvBinary(mx1->cols+t, mx2->cols+u, dst, fltLaNR)
; else
fn_minus(mx1->cols+t, mx2->cols+u, dst) /* compute t / u */
; if (overwrite)
mclvCopy(diff, dst)
, mclvCopy(dst, cache) /* reinstate column */
/* diff contains t / u */
; dst = overwrite ? dst : meet /* cache column, same as above */
; if (modes & MMM_BINARY)
mclvBinary(mx1->cols+t, mx2->cols+u, dst, fltLaR)
; else
fn_meet(mx1->cols+t, mx2->cols+u, dst)
; if (overwrite)
mclvCopy(meet, dst)
, mclvCopy(dst, cache) /* meet contains t /\ u */
; mcldMerge(diff, meet, join) /* join should be identical to column t */
; if (meet->n_ivps)
n_plus_meet++
; else
n_zero_meet++
; if (modes & MMM_CHECK)
{ mclv* dediff = mclvClone(mx1->cols+t)
; mclv* demeet = mclvClone(mx1->cols+t)
; dim nd = mclvUpdateMeet(dediff, diff, fltSubtract)
; dim nm = mclvUpdateMeet(demeet, meet, fltSubtract)
; if
( diff->n_ivps + meet->n_ivps != mx1->cols[t].n_ivps
|| !mcldEquate(join, mx1->cols+t, MCLD_EQT_EQUAL)
|| diff->n_ivps != nd
|| meet->n_ivps != nm
)
{ mclvaDump(mx1->cols+t, stdout, -1, " ", MCLVA_DUMP_HEADER_ON)
; mclvaDump(mx2->cols+u, stdout, -1, " ", MCLVA_DUMP_HEADER_ON)
; mclvaDump(meet, stdout, -1, " ", MCLVA_DUMP_HEADER_ON)
; mclvaDump(diff, stdout, -1, " ", MCLVA_DUMP_HEADER_ON)
; mcxDie(1, me, "rats")
; }
mclvFree(&dediff)
; mclvFree(&demeet)
; }
n_tst++
; }
}
fprintf
( stdout
, "meet was nonempty %.2f\n"
, (double) (n_plus_meet * 1.0f / n_tst)
)
; fprintf
( stdout
, "%d successful tests in %s%s %s mode (checked: %s)\n"
, (int) n_tst
, overwrite ? "overwrite" : "create"
, overwrite ? ( modes & MMM_RIGHT ? "-right" : "-left" ) : ""
, modes & MMM_BINARY
? "generic"
: "update"
, (modes & MMM_CHECK ? "yes" : "no")
)
; fprintf
( stdout
, "meet-can: %10lu\n"
"meet-zip: %10lu\n"
"meet-s/l: %10lu\n"
"diff-can: %10lu\n"
"diff-zip: %10lu\n"
"diff-s/l: %10lu\n"
, (ulong) nu_meet_can
, (ulong) nu_meet_zip
, (ulong) nu_meet_sl
, (ulong) nu_diff_can
, (ulong) nu_diff_zip
, (ulong) nu_diff_sl
)
; mclvFree(&cache)
; mclvFree(&meet)
; mclvFree(&join)
; mclvFree(&diff)
; }
static mclx* make_mx_from_pars
( mclxIOstreamer* streamer
, stream_state* iface
, void (*ivpmerge)(void* ivp1, const void* ivp2)
, mcxbits bits
)
{ mclpAR* pars = iface->pars
; long dc_max_seen = iface->map_c->max_seen
; long dr_max_seen = iface->map_r->max_seen
; mclx* mx = NULL
; mclv* domc, *domr
; dim i
; if (bits & MCLXIO_STREAM_235ANY)
{ if (streamer->cmax_235 > 0 && dc_max_seen < streamer->cmax_235 - 1)
dc_max_seen = streamer->cmax_235-1
; }
else if (bits & MCLXIO_STREAM_123)
{ if (streamer->cmax_123 > 0 && dc_max_seen < streamer->cmax_123 - 1)
dc_max_seen = streamer->cmax_123-1
; if (streamer->rmax_123 > 0 && dr_max_seen < streamer->rmax_123 - 1)
dr_max_seen = streamer->rmax_123-1
; }
mcxTell("stream", "maxc=%d maxr=%d", (int) dc_max_seen, (int) dr_max_seen)
; if (iface->pars_n_used != iface->map_c->max_seen+1)
mcxDie
( 1
, module
, "internal discrepancy: n_pars=%lu max_seen+1=%lu"
, (ulong) iface->pars_n_used
, (ulong) (iface->map_c->max_seen+1)
)
; if (dc_max_seen < 0 || dr_max_seen < 0)
{ if (dc_max_seen < -1 || dr_max_seen < -1)
{ mcxErr(module, "bad apples %ld %ld", dc_max_seen, dr_max_seen)
; return NULL
; }
else
mcxTell(module, "no assignments yield void/empty matrix")
; }
/* fixme: with extend and same tab, should still copy.
* then, there are still occasions where one would
* want to go the sparse route
*/
domc = iface->map_c->tab && (iface->bits & MCLXIO_STREAM_CTAB_RO)
? mclvClone(iface->map_c->tab->domain)
: mclvCanonical(NULL, dc_max_seen+1, 1.0)
; domr = iface->map_r->tab && (iface->bits & MCLXIO_STREAM_RTAB_RO)
? mclvClone(iface->map_r->tab->domain)
: mclvCanonical(NULL, dr_max_seen+1, 1.0)
; if (! (mx = mclxAllocZero(domc, domr)))
{ mclvFree(&domc)
; mclvFree(&domr)
; }
else
for (i=0;i<iface->pars_n_used;i++) /* careful with signedness */
{ long d = domc->ivps[i].idx
;if(DEBUG3)fprintf(stderr, "column %d alloc %d\n", (int) d, (int) iface->pars_n_alloc);
; mclvFromPAR(mx->cols+i, pars+d, 0, ivpmerge, NULL)
; }
return mx
; }
int main
( int argc
, const char* argv[]
)
{ mcxIO *xfin = mcxIOnew("-", "r")
; mcxIO *xfout = mcxIOnew("-", "w")
; mclMatrix *mx = NULL
; mclx* cmapx = NULL, *rmapx = NULL
; const char* me = "mcxmap"
; long cshift = 0
; long rshift = 0
; long cmul = 1
; long rmul = 1
; mcxIO* xf_cannc = NULL
; mcxIO* xf_cannr = NULL
; mcxstatus status = STATUS_OK
; mcxbool invert = FALSE
; mcxbool invertr = FALSE
; mcxbool invertc = FALSE
; mcxIO* xf_map_c = NULL, *xf_map_r = NULL, *xf_map = NULL, *xf_tab = NULL
; mcxOption* opts, *opt
; mcxstatus parseStatus = STATUS_OK
; mcxLogLevel =
MCX_LOG_AGGR | MCX_LOG_MODULE | MCX_LOG_IO | MCX_LOG_GAUGE | MCX_LOG_WARN
; mclxIOsetQMode("MCLXIOVERBOSITY", MCL_APP_VB_NO)
; mclx_app_init(stderr)
; mcxOptAnchorSortById(options, sizeof(options)/sizeof(mcxOptAnchor) -1)
; opts = mcxOptParse(options, (char**) argv, argc, 1, 0, &parseStatus)
; if (!opts)
exit(0)
; for (opt=opts;opt->anch;opt++)
{ mcxOptAnchor* anch = opt->anch
; switch(anch->id)
{ case MY_OPT_HELP
: case MY_OPT_APROPOS
: mcxOptApropos(stdout, me, syntax, 0, 0, options)
; return 0
;
case MY_OPT_VERSION
: app_report_version(me)
; return 0
;
case MY_OPT_IMX
: mcxIOnewName(xfin, opt->val)
; break
;
case MY_OPT_OUT
: mcxIOnewName(xfout, opt->val)
; break
;
case MY_OPT_MUL
: cmul = atol(opt->val)
; rmul = cmul
; break
;
case MY_OPT_CMUL
: cmul = atol(opt->val)
; break
;
case MY_OPT_RMUL
: rmul = atol(opt->val)
; break
;
case MY_OPT_SHIFT
: cshift = atol(opt->val)
; rshift = atol(opt->val)
; break
;
case MY_OPT_CSHIFT
: cshift = atol(opt->val)
; break
;
case MY_OPT_RSHIFT
: rshift = atol(opt->val)
; break
;
case MY_OPT_MAP
: xf_map = mcxIOnew(opt->val, "r")
; invert = FALSE
; break
;
case MY_OPT_CMAP
: invertc = FALSE
; xf_map_c = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_RMAP
: invertr = FALSE
; xf_map_r = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_MAPI
: invert = TRUE
; xf_map = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_CMAPI
: invertc = TRUE
; xf_map_c = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_RMAPI
: invertr = TRUE
; xf_map_r = mcxIOnew(opt->val, "r")
; break
;
case MY_OPT_MAKE_MAP
: xf_cannc = mcxIOnew(opt->val, "w")
; xf_cannr = xf_cannc
; break
;
case MY_OPT_MAKE_MAPC
: xf_cannc = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_MAKE_MAPR
: xf_cannr = mcxIOnew(opt->val, "w")
; break
;
case MY_OPT_TAB
: xf_tab = mcxIOnew(opt->val, "r")
; break
;
}
}
/* little special case. restructure when it grows */
if (xf_tab)
{ mclTab* tab1, *tab2
; if (xf_map)
{ mcxIOopen(xf_map, EXIT_ON_FAIL)
; cmapx = mclxRead(xf_map, EXIT_ON_FAIL)
; }
else
mcxDie(1, me, "-tab option requires -map option")
; tab1 = mclTabRead(xf_tab, NULL, EXIT_ON_FAIL)
; if ((tab2 = mclTabMap(tab1, cmapx)))
mclTabWrite(tab2, xfout, NULL, EXIT_ON_FAIL)
; else
mcxDie(1, me, "map file error (subsumption/bijection)")
; return 0
; }
mx = mclxRead(xfin, EXIT_ON_FAIL)
; if (xf_map)
{ mcxIOopen(xf_map, EXIT_ON_FAIL)
; cmapx = mclxRead(xf_map, EXIT_ON_FAIL)
; rmapx = cmapx
; }
else
{ if (xf_map_c)
{ mcxIOopen(xf_map_c, EXIT_ON_FAIL)
; cmapx = mclxRead(xf_map_c, EXIT_ON_FAIL)
; }
else if (cshift || cmul > 1)
cmapx
= mclxMakeMap
( mclvCopy(NULL, mx->dom_cols)
, mclvMap(NULL, cmul, cshift, mx->dom_cols)
)
; else if (xf_cannc) /* fixme slightly flaky interface */
{ cmapx
= mclxMakeMap
( mclvCopy(NULL, mx->dom_cols)
, mclvCanonical(NULL, mx->dom_cols->n_ivps, 1.0)
)
; mclxWrite(cmapx, xf_cannc, MCLXIO_VALUE_GETENV, RETURN_ON_FAIL)
; }
if (xf_map_r)
{ mcxIOopen(xf_map_r, EXIT_ON_FAIL)
; rmapx = mclxRead(xf_map_r, EXIT_ON_FAIL)
; }
else if (rshift || rmul > 1)
rmapx
= mclxMakeMap
( mclvCopy(NULL, mx->dom_rows)
, mclvMap(NULL, rmul, rshift, mx->dom_rows)
)
; else if (xf_cannr)
{ rmapx
= mclxMakeMap
( mclvCopy(NULL, mx->dom_rows)
, mclvCanonical(NULL, mx->dom_rows->n_ivps, 1.0)
)
; if (xf_cannr != xf_cannc)
mclxWrite(rmapx, xf_cannr, MCLXIO_VALUE_GETENV, RETURN_ON_FAIL)
; else if (!mclxIsGraph(mx))
mcxErr(me, "row map not written but matrix is not a graph")
; }
}
if (invert && cmapx && cmapx == rmapx)
{ mclx* cmapxi = mclxTranspose(cmapx)
; mclxFree(&cmapx)
; cmapx = rmapx = cmapxi
; }
else
{ if ((invert || invertr) && rmapx)
{ mclx* rmapxi = mclxTranspose(rmapx)
; mclxFree(&rmapx)
; rmapx = rmapxi
; }
if ((invert || invertc) && cmapx)
{ mclx* cmapxi = mclxTranspose(cmapx)
; mclxFree(&cmapx)
; cmapx = cmapxi
; }
}
; status = STATUS_FAIL
; do
{ if (cmapx && mclxMapCols(mx, cmapx))
break
; if (rmapx && mclxMapRows(mx, rmapx))
break
; status = STATUS_OK
; }
while (0)
; if (status)
{ mcxErr(me, "error, nothing written")
; return 1
; }
mclxWrite(mx, xfout, MCLXIO_VALUE_GETENV, EXIT_ON_FAIL)
; return 0
; }
mclx* mclxIOstreamIn
( mcxIO* xf
, mcxbits bits
, mclpAR* transform
, void (*ivpmerge)(void* ivp1, const void* ivp2)
, mclxIOstreamer* streamer
, mcxOnFail ON_FAIL
)
{ mcxstatus status = STATUS_FAIL
; const char* me = module
; mcxbool symmetric = bits & MCLXIO_STREAM_SYMMETRIC
; mcxbool mirror = bits & MCLXIO_STREAM_MIRROR
; mcxbool abc = bits & MCLXIO_STREAM_ABC ? TRUE : FALSE
; mcxbool one23 = bits & MCLXIO_STREAM_123 ? TRUE : FALSE
; mcxbool etc = bits & (MCLXIO_STREAM_ETC | MCLXIO_STREAM_ETC_AI) ? TRUE : FALSE
; mcxbool longlist = bits & (MCLXIO_STREAM_ETCANY | MCLXIO_STREAM_235ANY) ? TRUE : FALSE
; mcxTing* linebuf = mcxTingEmpty(NULL, 100)
; map_state map_c = { NULL, NULL, -1 , 0}
; map_state map_r = { NULL, NULL, -1 , 0}
; stream_state iface
; etc_state etcstate
; unsigned long n_ite = 0
; mclx* mx = NULL
; if (!ivpmerge)
ivpmerge = mclpMergeMax
; if (symmetric)
iface.map_c = &map_c /* this bit of hidgery-pokery */
, iface.map_r = &map_c /* is a crucial interfacummathingy */
; else
iface.map_c = &map_c
, iface.map_r = &map_r
;if(DEBUG2)fprintf(stderr, "%s abc\n", abc ? "yes" : "no")
; etcstate.etcbuf = NULL
; etcstate.etcbuf_ofs = 0
; etcstate.etcbuf_check = 0
; etcstate.x_prev = ULONG_MAX /* note we depend on ULONG_MAX + 1 == 0 */
; etcstate.n_y = 0
/* fixme incomplete and distributed initialization of iface */
; iface.pars = NULL
; iface.pars_n_alloc = 0
; iface.pars_n_used = 0
;if(DEBUG3)fprintf(stderr, "1 + max c %lu\n", (ulong) (iface.map_c->max_seen+1))
/* fixme: put the block below in a subroutine */
; while (1)
{ if (abc + one23 + longlist > TRUE) /* OUCH */
{ mcxErr(module, "multiple stream formats specified")
; break
; }
if (!symmetric && streamer->tab_sym_in)
{ mcxErr(module, "for now disallowed, single tab, different domains")
; break
; }
if ((!one23 && !abc && !longlist))
{ mcxErr(module, "not enough to get going")
; break
; }
/* These have maps associated with them.
* Note that bitsp may be changed (by filling in
* somewhat underspecified settings).
* todo hierverder: etc case supported below ?
*/
if (abc || etc)
stream_state_set_map(symmetric, &iface, streamer, &bits)
; if (xf->fp == NULL && (mcxIOopen(xf, ON_FAIL) != STATUS_OK))
{ mcxErr(me, "cannot open stream <%s>", xf->fn->str)
; break
; }
status = STATUS_OK
; break
; }
iface.bits = bits
; if (!status)
while (1)
{ unsigned long x = 876543210, y = 876543210
; double value = 0
; n_ite++
; iface.x = 0
; iface.y = 0
; if (n_ite % 20000 == 0)
fputc('.', stderr) /* fixme conditional to sth */
; if (n_ite % 1000000 == 0)
fprintf(stderr, " %ldM\n", (long) (n_ite / 1000000))
/*
* - the read routines largely manage iface, including
* map_c->max_seen and map_r->max_seen. It would be
* nice to encapsulate that management in a single
* place. Note the read_abc requirement that sometimes
* a label may need to be deleted from a hash. The fact
* that handle_label (called by read_etc and read_abc)
* also manages max_seen complicate encapsulation though.
*
* - read_etc manages its line buffer.
*/
; status
= one23 ? read_123(xf, linebuf, &iface, streamer, &value, bits)
: abc ? read_abc(xf, linebuf, &iface, &value)
: longlist ? read_etc(xf, &iface, &etcstate, &value)
: STATUS_FAIL
; x = iface.x
; y = iface.y
/* considerme: etc status ignore could still expand column range.
* do we change the status and deal with not incorporating the row,
* or do we keep status, and change realloc/ignore logic below?
*/
;if(0)fprintf(stderr, "#x now %lu status %s\n", (ulong) (iface.map_c->max_seen+1), MCXSTATUS(status))
/* etc/235 are special in that with NEW x and IGNORE y
* we respect x
* fixme: should not do that for auto-increment
*/
; if (status == STATUS_IGNORE) /* maybe restrict mode */
{ if
( longlist
&& iface.statusx == STATUS_NEW
&& iface.map_c->max_seen+1 > iface.pars_n_used /* note mixed-sign comparison */
)
{ if ((status = pars_realloc(&iface, iface.map_c->max_seen+1)))
break
; }
continue
; }
else if (status) /* FAIL or DONE */
break
; if
( iface.map_c->max_seen >= iface.pars_n_used /* note mixed-sign comparison */
&& (status = pars_realloc(&iface, iface.map_c->max_seen+1))
)
break
; status = STATUS_FAIL /* fixme restructure logic, mid-re-initialization is ugly */
; if
( bits & (MCLXIO_STREAM_LOGTRANSFORM | MCLXIO_STREAM_NEGLOGTRANSFORM) )
{ if (bits & MCLXIO_STREAM_LOGTRANSFORM)
value = value > 0 ? log(value) : -PVAL_MAX
; else if (bits & MCLXIO_STREAM_NEGLOGTRANSFORM)
value = value > 0 ? -log(value) : PVAL_MAX
; if (bits & MCLXIO_STREAM_LOG10)
value /= log(10)
; }
if (transform)
{ mclp bufivp
; bufivp.idx = 0
; bufivp.val = value
; value = mclpUnary(&bufivp, transform)
; }
/* fixme: below we have canonical dependence, index as offset */
if (value)
{ if(DEBUG3)fprintf(stderr, "attempt to extend %d\n", (int) x)
; if (mclpARextend(iface.pars+x, y, value))
{ mcxErr(me, "x-extend fails")
; break
; }
if (mirror && mclpARextend(iface.pars+y, x, value))
{ mcxErr(me, "y-extend fails")
; break
; }
}
status = STATUS_OK
; }
if (n_ite >= 1000000 && n_ite % 5000000)
fputc('\n', stderr)
; mcxTingFree(&(etcstate.etcbuf))
; if (status == STATUS_FAIL || ferror(xf->fp))
mcxErr(me, "error occurred (status %d lc %d)", (int) status, (int) xf->lc)
; else
{ mx = make_mx_from_pars(streamer, &iface, ivpmerge, bits)
; status = mx ? STATUS_OK : STATUS_FAIL
; }
mcxTingFree(&linebuf)
; free_pars(&iface)
; if (status == STATUS_FAIL)
{ if (ON_FAIL == EXIT_ON_FAIL)
mcxDie(1, me, "fini")
; }
/* with 123, etcai there is simply no column tab
* todo: perhaps create a dummy one (integers).
*/
if
( !status
&& (abc || (bits & (MCLXIO_STREAM_ETC | MCLXIO_STREAM_ETC_AI)))
)
{ if (symmetric)
streamer->tab_sym_out = make_tab(iface.map_c)
; else
{ if (!(bits & MCLXIO_STREAM_ETC_AI))
streamer->tab_col_out = make_tab(iface.map_c)
;if(0)fprintf(stderr, "%p x %p\n", (void*) iface.map_c->map, (void*) iface.map_c->tab)
;if(0)mcxHashStats(stdout, iface.map_c->map)
; streamer->tab_row_out = make_tab(iface.map_r)
; }
}
mcxHashFree(&(iface.map_c->map), mcxTingRelease, NULL)
; if (!symmetric)
mcxHashFree(&(iface.map_r->map), mcxTingRelease, NULL)
; return mx
; }
