+/*
+ */
+
+#include "structure.h"
+
+
+typedef struct {
+ int x, y;
+} Point;
+
+typedef struct { /* both inclusive */
+ Point tl;
+ Point br;
+} Rect;
+
+static CanonImage *cim;
+
+static inline char get(int x, int y) { return cim->d[y * cim->w + x]; }
+static inline char get_p(Point p) { return get(p.x,p.y); }
+
+#define START_MAIN {200,200}
+#define MIN_COLUMNS 6
+#define INTERESTING_COLUMNS 7
+#define TEXT_COLUMNS 2
+#define MAX_COLUMNS 7
+
+static Rect mainr = { START_MAIN,START_MAIN };
+static int commbasey, comminty;
+static int colrightx[INTERESTING_COLUMNS];
+static int text_h=-1, columns=-1;
+static OcrReader *rd;
+
+const CanonColourInfo canoncolourinfos[]= {
+ { 0x475A5E, '*' }, /* edge */
+ { 0x2C5F7A, '*' }, /* edge just under box heading shadow */
+ { 0x7D9094, '+' }, /* interbox */
+
+ { 0xBDC5BF, ' ' }, /* background - pale Sugar cane, etc. */
+ { 0xADB5AF, ' ' }, /* background - dark */
+ { 0xC7E1C3, ' ' }, /* background - pale Swill, etc. */
+ { 0xB5CFB1, ' ' }, /* background - dark */
+ { 0xD6CEB0, ' ' }, /* background - pale Madder, etc. */
+ { 0xC8C0A2, ' ' }, /* background - dark */
+ { 0xE0E1D3, ' ' }, /* background - pale Lorandite, etc. */
+ { 0xD0D1C3, ' ' }, /* background - dark */
+ { 0xE5E6C1, ' ' }, /* background - pale Cloth */
+ { 0xD7D8B3, ' ' }, /* background - dark */
+ { 0xEDDED9, ' ' }, /* background - pale Dye */
+ { 0xDACBC6, ' ' }, /* background - dark */
+ { 0xD3DEDF, ' ' }, /* background - pale Paint */
+ { 0xC5D0D1, ' ' }, /* background - dark */
+ { 0xDCD1CF, ' ' }, /* background - pale Enamel */
+ { 0xCEC3C1, ' ' }, /* background - dark */
+ { 0xF3F6F5, ' ' }, /* background - pale fruit */
+ { 0xE2E7E5, ' ' }, /* background - dark */
+
+ { 0x000000, 'o' }, /* foreground */
+ { 0xD4B356, ' ' }, /* background (cursor) */
+ { 0xFFFFFF, 'o' }, /* foreground (cursor) */
+
+ { 0x5B93BF, '_' }, /* selector dropdown background */
+ { 0xD7C94F, 'X' }, /* selector dropdown foreground */
+ { 0,0 }
+};
+
+static void require_rectangle(int tlx, int tly, int brx, int bry,
+ const char *ok) {
+ int x,y;
+ for (x=tlx; x<=brx; x++)
+ for (y=tly; y<=bry; y++) {
+ int c= get(x,y);
+ assert(strchr(ok,c));
+ }
+}
+static void require_rectangle_r(Rect rr, const char *ok) {
+ require_rectangle(rr.tl.x,rr.tl.y, rr.br.x,rr.br.y, ok);
+}
+
+static void debug_rect(const char *what, int whati, Rect rr) {
+#ifdef DEBUG_RECTANGLES
+ int y,r,w;
+ fprintf(debug, "%s %d: %d,%d..%d,%d:\n", what, whati,
+ rr.tl.x,rr.tl.y, rr.br.x,rr.br.y);
+ w= rr.br.x - rr.tl.x + 1;
+ for (y=rr.tl.y; y<=rr.br.y; y++) {
+ fprintf(debug, "%4d%*s|", y, rr.tl.x,"");
+ r= fwrite(cim->d + y*cim->w + rr.tl.x, 1, w, debug);
+ eassert(r==w);
+ fputc('|',debug);
+ fputc('\n',debug);
+ }
+#endif
+ debug_flush();
+}
+
+#define WALK_UNTIL(point,coord,increm,last,edge) \
+ for (;;) { \
+ if ((point).coord == (last)+(increm)) break; \
+ if (get_p((point)) == (edge)) { (point).coord -= (increm); break; } \
+ (point).coord += (increm); \
+ }
+
+#define WALK_UNTIL_MUST(point,coord,increm,last,edge) \
+ do { \
+ WALK_UNTIL(point,coord,increm,last,edge); \
+ eassert((point).coord != (last)+(increm)); \
+ } while(0)
+
+void find_structure(CanonImage *im) {
+ cim= im;
+
+ Rect whole = { {0,0}, {cim->w-1,cim->h-1} };
+
+ WALK_UNTIL_MUST(mainr.tl, x,-1, whole.tl.x, '*');
+ WALK_UNTIL_MUST(mainr.tl, y,-1, whole.tl.y, '*');
+ WALK_UNTIL_MUST(mainr.br, x,+1, whole.br.x, '*');
+ WALK_UNTIL_MUST(mainr.br, y,+1, whole.br.y, '*');
+
+ require_rectangle(mainr.tl.x-1, mainr.tl.y, mainr.tl.x-1, mainr.br.y, "*");
+ require_rectangle(mainr.br.x+1, mainr.tl.y, mainr.br.x+1, mainr.br.y, "*");
+ require_rectangle(mainr.tl.x, mainr.tl.y-1, mainr.br.x, mainr.tl.y-1, "*");
+ require_rectangle(mainr.tl.x, mainr.br.y+1, mainr.br.x, mainr.br.y+1, "*");
+
+#define CHECK_STRIP_BORDER(tlbr,xy,increm) \
+ do { \
+ Point csb_p; \
+ Rect csb_r; \
+ csb_p= mainr.tl; \
+ csb_p.xy= mainr.tlbr.xy; \
+ if (get_p(csb_p)=='+') { \
+ csb_r= mainr; \
+ csb_r.tl.xy= csb_p.xy; \
+ csb_r.br.xy= csb_p.xy; \
+ require_rectangle_r(csb_r, "+"); \
+ mainr.tlbr.xy += increm; \
+ } \
+ } while(0)
+
+ debug_rect("mainr",0, mainr);
+
+ CHECK_STRIP_BORDER(tl,x,+1);
+ CHECK_STRIP_BORDER(tl,y,+1);
+ CHECK_STRIP_BORDER(br,x,-1);
+ CHECK_STRIP_BORDER(br,y,-1);
+
+ debug_rect("mainr",1, mainr);
+
+ Point up = START_MAIN;
+ WALK_UNTIL_MUST(up, y,-1, mainr.tl.y, '+');
+
+ Point down = START_MAIN;
+ down.y++;
+ WALK_UNTIL_MUST(down, y,+1, mainr.br.y, '+');
+
+#ifdef DEBUG_RECTANGLES
+ int xscaleunit, y,x;
+ for (y=0, xscaleunit=1; y<4; y++, xscaleunit*=10) {
+ fprintf(debug," ");
+ for (x=0; x<=cim->w; x++) {
+ if (x % xscaleunit) fputc(' ',debug);
+ else fprintf(debug,"%d",(x / xscaleunit)%10);
+ }
+ fputc('\n',debug);
+ }
+#endif
+
+ commbasey= up.y;
+ comminty= down.y - up.y + 2;
+
+ Point across= { mainr.tl.x, commbasey };
+ int colno=0;
+ for (;;) {
+ eassert(get_p(across) != '+');
+ WALK_UNTIL(across, x,+1, mainr.br.x, '+');
+ eassert(colno < MAX_COLUMNS);
+ int colrx= across.x;
+ if (colrx > mainr.br.x) colrx= mainr.br.x;
+ if (colno < INTERESTING_COLUMNS)
+ colrightx[colno]= colrx;
+
+ colno++;
+
+ if (across.x >= mainr.br.x-1)
+ break;
+
+ across.x++;
+ require_rectangle(across.x,mainr.tl.y, across.x,mainr.br.y, "+");
+ across.x++;
+ }
+ eassert(colno >= MIN_COLUMNS);
+
+#define SET_ONCE(var,val) do{ \
+ int v= (val); \
+ if ((var)==-1) (var)= v; \
+ else eassert((var) == v); \
+ }while(0)
+
+ SET_ONCE(columns, colno);
+ SET_ONCE(text_h, comminty - 1);
+}
+
+CanonImage *alloc_canon_image(int w, int h) {
+ CanonImage *im= malloc(sizeof(CanonImage) + w*h);
+ eassert(im);
+ im->w= w;
+ im->h= h;
+ memset(im->d,'?',w*h);
+ return im;
+}
+
+CanonImage *file_read_image_ppm(FILE *f) {
+ struct pam inpam;
+ unsigned char rgb_buf[3];
+ CanonImage *im;
+
+ pnm_readpaminit(f, &inpam, sizeof(inpam));
+ eassert(inpam.maxval == 255);
+ eassert(inpam.bytes_per_sample == 1);
+
+ CANONICALISE_IMAGE(im, inpam.width, inpam.height, {
+ int r= fread(&rgb_buf,1,3,f); eassert(r==3);
+
+ rgb=
+ ((unsigned long)rgb_buf[0]<<16) |
+ ((unsigned long)rgb_buf[1]<<8) |
+ (rgb_buf[2]);
+ });
+
+ return im;
+}
+
+void read_screenshots(void) {
+ int c;
+ while ((c= fgetc(screenshots_file) != EOF)) {
+ ungetc(c, screenshots_file);
+
+ eassert(npages < MAX_PAGES);
+ page_images[npages++]= file_read_image_ppm(screenshots_file);
+ }
+}
+
+static void find_commodity(int offset, Rect *rr) {
+ /* rr->tl.x==-1 if offset out of range */
+ rr->tl.y= commbasey - offset*comminty;
+ rr->br.y= rr->tl.y + comminty-2;
+ if (rr->tl.y < mainr.tl.y || rr->br.y > mainr.br.y) { rr->tl.x=-1; return; }
+ if (rr->tl.y > mainr.tl.y)
+ require_rectangle(rr->tl.x,rr->tl.y-1, rr->br.x,rr->tl.y-1, "+");
+ if (rr->br.y < mainr.tl.y)
+ require_rectangle(rr->tl.x,rr->br.y+1, rr->br.x,rr->br.y+1, "+");
+
+ rr->tl.x= mainr.tl.x;
+ rr->br.x= mainr.br.x;
+}
+
+static void find_table_entry(Rect commod, int colno, Rect *cellr) {
+ cellr->tl.y= commod.tl.y;
+ cellr->br.y= commod.br.y;
+ cellr->tl.x= !colno ? commod.tl.x : colrightx[colno-1]+2;
+ cellr->br.x= colrightx[colno];
+ debug_rect("cell", colno, *cellr);
+ require_rectangle_r(*cellr, " o");
+}
+
+static void ocr_rectangle(Rect r, const OcrCellType ct) {
+ OcrResultGlyph *results, *res;
+
+ int w= r.br.x - r.tl.x + 1;
+ Pixcol cols[w+1];
+ int x,y;
+ for (x=0; x<w; x++) {
+ Pixcol cx, rv;
+ for (y=0, cx=0, rv=1; y<text_h; y++, rv<<=1) {
+ switch (get(x+r.tl.x, y+r.tl.y)) {
+ case ' ': break;
+ case 'o': cx |= rv; break;
+ default: eassert(!"wrong pixel");
+ }
+ }
+ cols[x]= cx;
+ }
+ cols[w]= 0;
+
+ results= ocr(rd,ct,w,cols);
+ for (res=results; res->s; res++)
+ printf("%s",res->s);
+}
+
+void analyse(void) {
+ Rect thisr, entryr;
+ int page, tryrect, colno;
+
+ for (page=0; page<npages; page++) {
+ find_structure(page_images[page]);
+
+ if (!rd)
+ rd= ocr_init(text_h);
+
+ for (tryrect= +cim->h; tryrect >= -cim->h; tryrect--) {
+ find_commodity(tryrect, &thisr);
+ if (thisr.tl.x < 0)
+ continue;
+ debug_rect("commod",tryrect, thisr);
+
+ const char *tab= "";
+ for (colno=0; colno<columns; colno++) {
+ find_table_entry(thisr,colno,&entryr);
+ fputs(tab, stdout);
+ ocr_rectangle(entryr,
+ colno<TEXT_COLUMNS
+ ? &ocr_celltype_text
+ : &ocr_celltype_number);
+ tab= "\t";
+ }
+ fputs("\n", stdout);
+ eassert(!ferror(stdout));
+ eassert(!fflush(stdout));
+ }
+ }
+}