From: Ian Jackson Date: Sun, 19 Jul 2009 16:35:21 +0000 (+0100) Subject: Merge branch 'origin' X-Git-Tag: 3.0~44^2~1 X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?p=ypp-sc-tools.web-live.git;a=commitdiff_plain;h=b3a5bc5d04aa5c9d52f3987965fdb2a4cd4c8e14;hp=609d1bc79abd962d7722607d5dda7c202db9497d Merge branch 'origin' --- diff --git a/pctb/common.h b/pctb/common.h index 1fb7e11..9d06fab 100644 --- a/pctb/common.h +++ b/pctb/common.h @@ -53,7 +53,6 @@ typedef struct { int w,h; - const struct RgbImage *rgb; char d[]; } CanonImage; diff --git a/pctb/convert.c b/pctb/convert.c index 4fdb063..1f13850 100644 --- a/pctb/convert.c +++ b/pctb/convert.c @@ -36,28 +36,12 @@ const char *get_vardir(void) { return "."; } const char *get_libdir(void) { return "."; } -enum mode { - mf_findwindow= 00001, - mf_screenshot= 00010, - mf_readscreenshot= 00020, - mf_analyse= 00100, - mfm_special= 07000, - - mode_findwindow= 00001, - mode_screenshot= 00011, - mode_analyse= 00120, - mode_showcharset= 01000, - - mode_all= 00111, -}; - enum outmodekind { omk_unset, omk_upload, omk_str, omk_raw, omk_none }; static enum outmodekind o_outmode_kind; static const char *o_outmode_str= 0; -static enum mode o_mode= mode_all; static char *o_screenshot_fn; static const char *o_serv_pctb, *o_serv_dict_fetch, *o_serv_dict_submit; @@ -68,6 +52,7 @@ int o_quiet; static pid_t screenshot_compressor=-1; +enum mode o_mode= mode_all; enum flags o_flags= ff_charset_allowedit | ff_dict_fetch|ff_dict_submit|ff_dict_pirate; diff --git a/pctb/convert.h b/pctb/convert.h index 0dc2255..11f534b 100644 --- a/pctb/convert.h +++ b/pctb/convert.h @@ -42,37 +42,41 @@ #define MAXIMGIDENT 100 +typedef uint32_t Rgb; /* red<<0 | green<<8 | blue<<16 */ + typedef struct RgbImage { int w, h; - unsigned char data[]; - /* red = data[ y*w*3 + x*3 + 0 ] = RI_PIXEL(ri,x,y)[0] - * green = data[ y*w*3 + x*3 + 1 ] = RI_PIXEL(ri,x,y)[1] - * blue = data[ y*w*3 + x*3 + 2 ] = RI_PIXEL(ri,x,y)[2] - */ + Rgb data[]; /* data[ y*w + x ] */ } RgbImage; -typedef unsigned long Rgb; - void identify_rgbimage(const RgbImage *base, Rect portion, char result[MAXIMGIDENT], const char *what); RgbImage *alloc_rgb_image(int w, int h); void fwrite_ppmraw(FILE *f, const RgbImage *ri); -#define RI_PIXEL(ri,x,y) ((ri)->data + ((y)*(ri)->w + (x)) * 3) +#define RI_PIXEL32(ri,x,y) ((ri)->data + ((y)*(ri)->w + (x))) static inline Rgb ri_rgb(const RgbImage *ri, int x, int y) { - const unsigned char *rip= RI_PIXEL(ri,x,y); - return (rip[0] << 16) | - (rip[1] << 8) | - (rip[2] ); + return *RI_PIXEL32(ri,x,y); } /*----- from structure.c -----*/ -void find_structure(CanonImage *im, int *max_relevant_y_r, +typedef struct PageStruct PageStruct; + +void find_structure(const CanonImage *im, + PageStruct **pagestruct_r, /* caller must free() */ + int *max_relevant_y_r, Point *commod_focus_point_r, Point *commod_page_point_r, Point *commod_focuslast_point_r); +void store_current_page(CanonImage *ci /*pointer saved*/, + PageStruct *pagestruct, /*freed*/ + RgbImage *rgb /*freed*/); +void adjust_colours(CanonImage *ci, const RgbImage *rgb); + +void select_page(int page); + Rect find_sunshine_widget(void); void canon_colour_prepare(void); @@ -122,6 +126,22 @@ enum flags { }; extern enum flags o_flags; +enum mode { + mf_findwindow= 00001, + mf_screenshot= 00010, + mf_readscreenshot= 00020, + mf_analyse= 00100, + mfm_special= 07000, + + mode_findwindow= 00001, + mode_screenshot= 00011, + mode_analyse= 00120, + mode_showcharset= 01000, + + mode_all= 00111, +}; +extern enum mode o_mode; + extern const char *o_ocean, *o_pirate; extern int o_quiet; @@ -134,7 +154,9 @@ void take_screenshots(void); void take_one_screenshot(void); #define MAX_PAGES 1000 -extern CanonImage *page_images[MAX_PAGES]; + +extern const CanonImage *page_images[MAX_PAGES]; +extern const RgbImage *page0_rgbimage; extern int npages; extern const char *ocean, *pirate; diff --git a/pctb/ocr.c b/pctb/ocr.c index f7cb450..92b0903 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -397,7 +397,7 @@ static DatabaseNode *findchar_1ctx(const FindCharArgs *fca, int x= fca->x; for (;;) { - debug_flush(); + if (DEBUGP(ocr)) debug_flush(); debugf(" | x=%d",x); if (x > fca->w) break; Pixcol cv= fca->cols[x]; @@ -531,7 +531,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, x=0; for (;;) { - debug_flush(); + if (DEBUGP(ocr)) debug_flush(); /* skip spaces */ if (x>=w) break; diff --git a/pctb/pages.c b/pctb/pages.c index 9e7a2b2..2e3edde 100644 --- a/pctb/pages.c +++ b/pctb/pages.c @@ -41,9 +41,6 @@ #include #include -CanonImage *page_images[MAX_PAGES]; -int npages; - const char *ocean, *pirate; static XWindowAttributes attr; @@ -267,9 +264,9 @@ static void snapshot(Snapshot **output) { #define COMPUTE_SHIFT_MASK(ix, rgb) \ compute_shift_mask(&shiftmasks[ix], im_use->rgb##_mask) - COMPUTE_SHIFT_MASK(0, red); + COMPUTE_SHIFT_MASK(0, blue); COMPUTE_SHIFT_MASK(1, green); - COMPUTE_SHIFT_MASK(2, blue); + COMPUTE_SHIFT_MASK(2, red); if (!*output) *output= alloc_rgb_image(wwidth, wheight); @@ -277,7 +274,7 @@ static void snapshot(Snapshot **output) { rtimestamp(&begin, "compute_shift_masks+alloc_rgb_image"); int x,y,i; - unsigned char *op= (*output)->data; + uint32_t *op32= (*output)->data; for (y=0; yxoffset == 0 && im_use->format == ZPixmap && @@ -289,21 +286,19 @@ static void snapshot(Snapshot **output) { im_use->blue_mask == 0xff0000U) { const char *p= im_use->data + y * im_use->bytes_per_line; // debugf("optimised copy y=%d",y); - for (x=0; x> shiftmasks[i].rshift) & SAMPLEMASK; - *op++= sample; } + *op32++= sample; } } } @@ -355,7 +350,7 @@ static void wait_for_stability(Snapshot **output, " last_input=%f previously=%p `%s'\n", last_input, previously, doing); - double min_interval= 0.025; /*us*/ + double min_interval= 0.025; for (;;) { progress_spinner("%s",doing); @@ -520,21 +515,23 @@ static void set_focus_commodity(void) { debugf("PAGING raise_and_set_focus done.\n"); } -static CanonImage *convert_page(const Snapshot *sn) { +static CanonImage *convert_page(const Snapshot *sn, RgbImage **rgb_r) { CanonImage *im; + RgbImage *ri; fwrite_ppmraw(screenshot_file, sn); - const unsigned char *pixel= sn->data; - CANONICALISE_IMAGE(im, sn->w, sn->h, { - r= *pixel++; - g= *pixel++; - b= *pixel++; + const Rgb *pixel= sn->data; + CANONICALISE_IMAGE(im, sn->w, sn->h, ri, { + rgb= *pixel++; }); - + sysassert(!ferror(screenshot_file)); sysassert(!fflush(screenshot_file)); + if (rgb_r) *rgb_r= ri; + else free(ri); + return im; } @@ -546,8 +543,8 @@ static void prepare_ypp_client(void) { raise_and_get_details(); wait_for_stability(¤t,0,0, "checking current YPP client screen..."); - test= convert_page(current); - find_structure(test, &max_relevant_y, + test= convert_page(current,0); + find_structure(test,0, &max_relevant_y, &commod_focus_point, &commod_page_point, &commod_focuslast_point); @@ -577,6 +574,20 @@ static void prepare_ypp_client(void) { free_snapshot(&status); } +static void convert_store_page(Snapshot *current) { + RgbImage *rgb; + CanonImage *ci; + PageStruct *pstruct; + + progress("page %d prescanning ...",npages); + ci= convert_page(current,&rgb); + + progress("page %d overview ...",npages); + find_structure(ci,&pstruct, 0,0,0,0); + + store_current_page(ci,pstruct,rgb); +} + void take_screenshots(void) { Snapshot *current=0, *last=0; @@ -589,19 +600,19 @@ void take_screenshots(void) { /* now to actually page down */ for (;;) { - debugf("paging page %d\n",npages); + debugf("page %d paging\n",npages); if (!(npages < MAX_PAGES)) fatal("Paging down seems to generate too many pages - max is %d.", MAX_PAGES); - - page_images[npages]= convert_page(current); + + convert_store_page(current); free_snapshot(&last); last=current; current=0; debugf("PAGING page %d converted\n",npages); wait_for_stability(¤t,last, 0, - "collecting screenshot of page %d...", + "page %d collecting ...", npages+1); if (npages && /* first pagedown doesn't do much */ @@ -629,7 +640,7 @@ void take_one_screenshot(void) { prepare_ypp_client(); wait_for_stability(¤t,0,0, "taking screenshot..."); - page_images[0]= convert_page(current); + convert_store_page(current); npages= 1; progress_log("collected single screenshot."); } diff --git a/pctb/rgbimage.c b/pctb/rgbimage.c index 81eddf9..dc9fa68 100644 --- a/pctb/rgbimage.c +++ b/pctb/rgbimage.c @@ -85,14 +85,16 @@ static int identify1(const RgbImage *base, Rect portion, int x,y,i; for (y=0; y=0 && c<=255); - int px= portion.tl.x + x, py= portion.tl.y + y; - diff |= px > portion.br.x || py > portion.br.y || - (c != RI_PIXEL(base,px,py)[i]); + rgb |= (Rgb)c << (i*8); } + int px= portion.tl.x + x, py= portion.tl.y + y; + diff |= px > portion.br.x || py > portion.br.y || + rgb != ri_rgb(base,px,py); } } if (!diff) { @@ -116,12 +118,17 @@ static int identify(const RgbImage *base, Rect portion, } void fwrite_ppmraw(FILE *f, const RgbImage *ri) { + int i; fprintf(f, "P6\n" "%d %d\n" "255\n", ri->w, ri->h); - int count= ri->w * ri->h * 3; - sysassert( fwrite(ri->data, 1, count, f) == count ); + for (i=0; i < ri->w * ri->h; i++) { + Rgb rgb= ri->data[i]; + fputc_unlocked(rgb >> 0, f); + fputc_unlocked(rgb >> 8, f); + fputc_unlocked(rgb >> 16, f); + } sysassert(!ferror(f)); sysassert(!fflush(f)); } @@ -132,8 +139,9 @@ static void fwrite_ppm(FILE *f, const RgbImage *base, Rect portion) { for (y=portion.tl.y; y<=portion.br.y; y++) { for (x=portion.tl.x; x<=portion.br.x; x++) { putc(' ',f); + Rgb rgb= ri_rgb(base,x,y); for (i=0; i<3; i++) - fprintf(f," %3d", RI_PIXEL(base,x,y)[i]); + fprintf(f," %3d", (rgb>>(i*8)) & 0xff); } putc('\n',f); } @@ -173,7 +181,7 @@ void identify_rgbimage(const RgbImage *base, Rect portion, RgbImage *alloc_rgb_image(int w, int h) { RgbImage *ri; - ri= mmalloc(sizeof(*ri) + w*h*3); + ri= mmalloc(sizeof(*ri) + w*h*sizeof(ri->data[0])); ri->w= w; ri->h= h; return ri; diff --git a/pctb/structure.c b/pctb/structure.c index ad95cb5..f10addc 100644 --- a/pctb/structure.c +++ b/pctb/structure.c @@ -27,11 +27,6 @@ #include "structure.h" -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); } - DEBUG_DEFINE_DEBUGF(struct) #define START_MAIN {200,200} @@ -40,19 +35,39 @@ DEBUG_DEFINE_DEBUGF(struct) #define TEXT_COLUMNS 2 #define MAX_COLUMNS 7 -static Rect mainr; -static int commbasey, comminty; -static int colrightx[INTERESTING_COLUMNS]; +struct PageStruct { + Rect mr; + int commbasey, comminty; + int colrightx[INTERESTING_COLUMNS]; +}; + +const CanonImage *page_images[MAX_PAGES]; +static PageStruct page_structs[MAX_PAGES]; +const RgbImage *page0_rgbimage; +int npages; + static int text_h=-1, columns=-1; + static OcrReader *rd; +static const CanonImage *cim; +static PageStruct s; + char *archipelago, *island; #define OTHERCOORD_x y #define OTHERCOORD_y x + +void select_page(int page) { + cim= page_images[page]; + s= page_structs[page]; + assert(cim); +} + + typedef struct { - Rgb rgb; /* on screen */ + Rgb rgbx; /* on screen, REVERSED BYTES ie r||g||b */ char c; /* canonical */ } CanonColourInfo; @@ -99,9 +114,9 @@ CanonColourInfoReds canoncolourinfo_tree; void canon_colour_prepare(void) { const CanonColourInfo *cci; for (cci=canoncolourinfo_table; cci->c; cci++) { - unsigned char r= cci->rgb >> 16; - unsigned char g= cci->rgb >> 8; - unsigned char b= cci->rgb; + unsigned char r= cci->rgbx >> 16; + unsigned char g= cci->rgbx >> 8; + unsigned char b= cci->rgbx; CanonColourInfoGreens *greens= canoncolourinfo_tree.red2[r]; if (!greens) { @@ -119,6 +134,9 @@ void canon_colour_prepare(void) { } } +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); } + static void mustfail1(const char *file, int line, const char *what) { fprintf(stderr, @@ -142,7 +160,7 @@ static void mustfail2(void) { } #define MUST(x, ifnot) do{ \ - if (!(x)) { \ + if (__builtin_expect(!(x), 0)) { \ mustfail1(__FILE__,__LINE__,#x); \ ifnot; \ mustfail2(); \ @@ -152,7 +170,7 @@ static void mustfail2(void) { #define MP(v) fprintf(stderr," %s=%d,%d",#v,(v).x,(v).y) #define MI(v) fprintf(stderr," %s=%d", #v,(v)) #define MIL(v) fprintf(stderr," %s=%ld", #v,(v)) -#define MRGB(v) fprintf(stderr," %s=%06lx", #v,(v)) +#define MRGB(v) fprintf(stderr," %s=%06"PRIx32, #v,(v)) #define MC(v) fprintf(stderr," %s='%c'", #v,(v)) #define MS(v) fprintf(stderr," %s=\"%s\"", #v,(v)) #define MF(v) fprintf(stderr," %s=%f", #v,(v)) @@ -255,12 +273,15 @@ static int commod_selector_matches(Rect search, const char *const *all, (search).TLBR.XY += increm; \ } -void find_structure(CanonImage *im, int *max_relevant_y_r, +void find_structure(const CanonImage *im, + PageStruct **pagestruct_r, + int *max_relevant_y_r, Point *commod_focus_point_r, Point *commod_page_point_r, Point *commod_focuslast_point_r) { cim= im; - + + FILLZERO(s); Rect whole = { {0,0}, {cim->w-1,cim->h-1} }; if (DEBUGP(rect)) { @@ -276,44 +297,44 @@ void find_structure(CanonImage *im, int *max_relevant_y_r, } Point mainr_tl= START_MAIN; - mainr.tl= mainr_tl; - WALK_UNTIL_MUST(mainr.tl, y,-1, whole.tl.y, ' '); - mainr.br= mainr.tl; + s.mr.tl= mainr_tl; + WALK_UNTIL_MUST(s.mr.tl, y,-1, whole.tl.y, ' '); + s.mr.br= s.mr.tl; - 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, '*'); + WALK_UNTIL_MUST(s.mr.tl, x,-1, whole.tl.x, '*'); + WALK_UNTIL_MUST(s.mr.tl, y,-1, whole.tl.y, '*'); + WALK_UNTIL_MUST(s.mr.br, x,+1, whole.br.x, '*'); + WALK_UNTIL_MUST(s.mr.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, "*"); + REQUIRE_RECTANGLE(s.mr.tl.x-1, s.mr.tl.y, s.mr.tl.x-1, s.mr.br.y, "*"); + REQUIRE_RECTANGLE(s.mr.br.x+1, s.mr.tl.y, s.mr.br.x+1, s.mr.br.y, "*"); + REQUIRE_RECTANGLE(s.mr.tl.x, s.mr.tl.y-1, s.mr.br.x, s.mr.tl.y-1, "*"); + REQUIRE_RECTANGLE(s.mr.tl.x, s.mr.br.y+1, s.mr.br.x, s.mr.br.y+1, "*"); #define CHECK_STRIP_BORDER(tlbr,xy,increm) \ do { \ Point csb_p; \ Rect csb_r; \ - csb_p= mainr.tl; \ + csb_p= s.mr.tl; \ csb_p.x++; csb_p.y++; \ - csb_p.xy= mainr.tlbr.xy; \ + csb_p.xy= s.mr.tlbr.xy; \ if (get_p(csb_p)=='+') { \ - csb_r= mainr; \ + csb_r= s.mr; \ csb_r.tl.xy= csb_p.xy; \ csb_r.br.xy= csb_p.xy; \ require_rectangle_r(csb_r, "+", __LINE__); \ - mainr.tlbr.xy += increm; \ + s.mr.tlbr.xy += increm; \ } \ } while(0) - debug_rect("mainr",0, mainr); + debug_rect("s.mr",0, s.mr); 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); + debug_rect("s.mr",1, s.mr); Rect updown= {START_MAIN,START_MAIN}; const int chkw= 100; @@ -321,46 +342,47 @@ void find_structure(CanonImage *im, int *max_relevant_y_r, updown.br.y++; debug_rect("updown",__LINE__,updown); - ADJUST_BOX(updown, "+", >=,chkw, mainr.tl.y, MUST, tl,y,-1); + ADJUST_BOX(updown, "+", >=,chkw, s.mr.tl.y, MUST, tl,y,-1); debug_rect("updown",__LINE__,updown); updown.br.y= updown.tl.y; updown.tl.y= updown.tl.y-1; - ADJUST_BOX(updown, "+*",>=,chkw, mainr.tl.y-1, MUST, tl,y,-1); + ADJUST_BOX(updown, "+*",>=,chkw, s.mr.tl.y-1, MUST, tl,y,-1); debug_rect("updown",__LINE__,updown); - commbasey= updown.tl.y + 1; - comminty= updown.br.y - updown.tl.y; + s.commbasey= updown.tl.y + 1; + s.comminty= updown.br.y - updown.tl.y; - Rect across= {{ mainr.tl.x - 1, commbasey }, - { mainr.tl.x, commbasey + comminty-2 }}; + Rect across= {{ s.mr.tl.x - 1, s.commbasey }, + { s.mr.tl.x, s.commbasey + s.comminty-2 }}; int colno=0; for (;;) { #define LIMIT_QUITEQ(cond,mp) { if (!(cond)) break; } debug_rect("across",colno*1000000+__LINE__, across); - ADJUST_BOX(across, "+",>=,comminty-1, mainr.br.x, LIMIT_QUITEQ, br,x,+1); + ADJUST_BOX(across, "+",>=,s.comminty-1,s.mr.br.x,LIMIT_QUITEQ,br,x,+1); debug_rect("across",colno*1000000+__LINE__, across); - MUST( colno < MAX_COLUMNS, MI(colno);MR(across);MR(mainr);MI(commbasey); ); + MUST( colno < MAX_COLUMNS, + MI(colno);MR(across);MR(s.mr);MI(s.commbasey); ); int colrx= across.br.x-1; - if (colrx >= mainr.br.x) colrx= mainr.br.x; + if (colrx >= s.mr.br.x) colrx= s.mr.br.x; if (colno < INTERESTING_COLUMNS) - colrightx[colno]= colrx; + s.colrightx[colno]= colrx; colno++; - if (across.br.x >= mainr.br.x) + if (across.br.x >= s.mr.br.x) break; - REQUIRE_RECTANGLE(across.br.x,mainr.tl.y, across.br.x,mainr.br.y, "+"); + REQUIRE_RECTANGLE(across.br.x,s.mr.tl.y, across.br.x,s.mr.br.y, "+"); across.br.x++; } - MUST( colno >= MIN_COLUMNS, MI(colno);MR(mainr);MR(across); ); + MUST( colno >= MIN_COLUMNS, MI(colno);MR(s.mr);MR(across); ); const int pagerh= 6; - Rect pager= {{ mainr.br.x, mainr.br.y - (pagerh-1) }, - { mainr.br.x + 1, mainr.br.y }}; + Rect pager= {{ s.mr.br.x, s.mr.br.y - (pagerh-1) }, + { s.mr.br.x + 1, s.mr.br.y }}; debug_rect("pager",__LINE__,pager); ADJUST_BOX(pager, "o",>=,pagerh-2, whole.br.x,MUST, br,x,+1); @@ -372,29 +394,35 @@ void find_structure(CanonImage *im, int *max_relevant_y_r, ADJUST_BOX(pager, "o",>=,pagerh-2, whole.br.x,MUST, br,x,+1); debug_rect("pager",__LINE__,pager); - ADJUST_BOX(pager, "o",>=,RECT_W(pager)-2, mainr.tl.y,LIMIT_QUITEQ, tl,y,-1); + ADJUST_BOX(pager, "o",>=,RECT_W(pager)-2, s.mr.tl.y,LIMIT_QUITEQ, tl,y,-1); debug_rect("pager",__LINE__,pager); #define SET_ONCE(var,val) do{ \ int v= (val); \ if ((var)==-1) (var)= v; \ - else MUST( (var) == v, MSB(#var);MI((var));MI(v);MR(mainr); ); \ + else MUST( (var) == v, MSB(#var);MI((var));MI(v);MR(s.mr); ); \ }while(0) SET_ONCE(columns, colno); - SET_ONCE(text_h, comminty - 1); + SET_ONCE(text_h, s.comminty - 1); + + if (pagestruct_r) { + *pagestruct_r= mmalloc(sizeof(s)); + **pagestruct_r= s; + } + if (max_relevant_y_r) - SET_ONCE(*max_relevant_y_r, mainr.br.y + 10); + SET_ONCE(*max_relevant_y_r, s.mr.br.y + 10); if (commod_focus_point_r) { - *commod_focus_point_r= mainr.tl; + *commod_focus_point_r= s.mr.tl; commod_focus_point_r->x += 10; - commod_focus_point_r->y += comminty/3; + commod_focus_point_r->y += s.comminty/3; } if (commod_focuslast_point_r) { - *commod_focuslast_point_r= mainr.br; + *commod_focuslast_point_r= s.mr.br; commod_focuslast_point_r->x -= 10; - commod_focuslast_point_r->y -= comminty/3; + commod_focuslast_point_r->y -= s.comminty/3; } if (commod_page_point_r) { commod_page_point_r->x= (pager.tl.x + pager.br.x) / 2; @@ -463,6 +491,10 @@ static void file_read_image_ppm(FILE *f) { struct pam inpam; unsigned char rgb_buf[3]; CanonImage *im; + RgbImage *ri; + PageStruct *pstruct; + + progress("page %d reading ...",npages); pnm_readpaminit(f, &inpam, sizeof(inpam)); if (!(inpam.maxval == 255 && @@ -470,14 +502,12 @@ static void file_read_image_ppm(FILE *f) { inpam.format == RPPM_FORMAT)) fatal("PNM screenshot(s) file must be 8bpp 1 byte-per-sample RGB raw"); - CANONICALISE_IMAGE(im, inpam.width, inpam.height, { - int rr= fread(&rgb_buf,1,3,f); - sysassert(!ferror(f)); + CANONICALISE_IMAGE(im, inpam.width, inpam.height, ri, { + errno=0; int rr= fread_unlocked(&rgb_buf,1,3,f); + sysassert(rr==3); if (rr!=3) fatal("PNM screenshot(s) file ends unexpectedly"); - r= rgb_buf[0]; - g= rgb_buf[1]; - b= rgb_buf[2]; + rgb= rgb_buf[0] | (rgb_buf[1] << 8) | (rgb_buf[2] << 16); }); sysassert(!ferror(screenshot_file)); @@ -485,7 +515,21 @@ static void file_read_image_ppm(FILE *f) { if (!(npages < MAX_PAGES)) fatal("Too many images in screenshots file; max is %d.\n", MAX_PAGES); - page_images[npages++]= im; + find_structure(im,&pstruct, 0,0,0,0); + store_current_page(im,pstruct,ri); + npages++; +} + +void store_current_page(CanonImage *ci, PageStruct *pstruct, RgbImage *rgb) { + assert(ci==cim); + progress("page %d unantialiasing...",npages); + adjust_colours(ci, rgb); + progress("page %d storing ...",npages); + if (!npages) page0_rgbimage= rgb; + else free(rgb); + page_images[npages]= cim; + page_structs[npages]= *pstruct; + free(pstruct); } void read_one_screenshot(void) { @@ -508,97 +552,183 @@ void read_screenshots(void) { if (c==EOF) break; ungetc(c, screenshot_file); } - progress("reading screenshot %d...",npages); file_read_image_ppm(screenshot_file); } sysassert(!ferror(screenshot_file)); progress_log("read %d screenshots.",npages); } -static double find_aa_density(const RgbImage *ri, Point p, long background, - long foreground, int fg_extra) { +#define FIXPT_SHIFT 15 + +typedef long Fixpt; +static inline Fixpt int2fixpt(int x) { return x<> (i*8)) & 0xff ); + fg_chan[i]= aa_foreground >> (i*8); + + aa_scale_chan[i]= 1.0 / (int2fixpt(fg_chan[i]) + fg_extra - aa_bg_chan[i]) + * dbl2fixpt(1) * dbl2fixpt(1); + } +} + +static inline Fixpt find_aa_density(const RgbImage *ri, Point p) { Rgb here= ri_rgb(ri, p.x, p.y); - double alpha[3], alpha_mean=0; + if (here==aa_background) return 0; + + Fixpt alpha[3], alpha_total=0; int i; for (i=0; i<3; i++) { - unsigned char here_chan= here >> (i*8); - unsigned char bg_chan= background >> (i*8); - unsigned char fg_chan= foreground >> (i*8); - double alpha_chan= - ((double)here_chan - (double)bg_chan) / - ((fg_chan + fg_extra) - (double)bg_chan); + unsigned char here_chan= here >> (i*8); + + Fixpt alpha_chan= fixpt_mul(int2fixpt(here_chan) - aa_bg_chan[i], + aa_scale_chan[i]); alpha[i]= alpha_chan; - alpha_mean += alpha_chan * (1/3.0); + alpha_total += alpha_chan; } - double thresh= 1.5/AAMAXVAL; - double alpha_min= alpha_mean - thresh; - double alpha_max= alpha_mean + thresh; + Fixpt one_third= dbl2fixpt(1/3.0); + Fixpt alpha_mean= fixpt_mul(alpha_total, one_third); + + Fixpt thresh= dbl2fixpt(1.5/AAMAXVAL); + Fixpt alpha_min= alpha_mean - thresh*2; + Fixpt alpha_max= alpha_mean + thresh*2; + for (i=0; i<3; i++) MUST( alpha_min <= alpha[i] && alpha[i] <= alpha_max, MP(p); - MRGB(here);MRGB(background);MRGB(foreground);MI(fg_extra); - MF(alpha_min); MI(i);MF(alpha[i]);MF(alpha_max) ); - - if ( -1e-5 < alpha_mean && alpha_mean <= 0.0 ) alpha_mean= 0.0; - if (1.0 <= alpha_mean && alpha_mean <= 1.0+1e-5) alpha_mean= 1.0; + MRGB(here);MRGB(aa_background);MRGB(aa_foreground); + MFP(aa_alpha_mean_max);MFP(thresh);MFP(alpha_mean); + MFP(alpha_min);MI(i);MFP(alpha[i]);MFP(alpha_max) ); - MUST( 0 <= alpha_mean && - (fg_extra ? (alpha_mean < 0.999) : (alpha_mean <= 1.0)), + MUST( -thresh <= alpha_mean && alpha_mean <= aa_alpha_mean_max + thresh, MP(p); - MRGB(here);MRGB(background);MRGB(foreground);MI(fg_extra); - MF(alpha_mean); MF(alpha[0]);MF(alpha[1]);MF(alpha[2]); ); + MRGB(here);MRGB(aa_background);MRGB(aa_foreground); + MFP(aa_alpha_mean_max);MFP(thresh); + MFP(alpha_mean); MFP(alpha[0]);MFP(alpha[1]);MFP(alpha[2]); ); + + if (alpha_mean < 0) alpha_mean= 0; + if (alpha_mean > aa_alpha_mean_max) alpha_mean= aa_alpha_mean_max; return alpha_mean; } 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; } + rr->tl.y= s.commbasey - offset*s.comminty; + rr->br.y= rr->tl.y + s.comminty-2; + if (rr->tl.y < s.mr.tl.y || rr->br.y > s.mr.br.y) { rr->tl.x=-1; return; } - rr->tl.x= mainr.tl.x; - rr->br.x= mainr.br.x; + rr->tl.x= s.mr.tl.x; + rr->br.x= s.mr.br.x; - if (rr->tl.y > mainr.tl.y) + if (rr->tl.y > s.mr.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) + if (rr->br.y < s.mr.tl.y) REQUIRE_RECTANGLE(rr->tl.x,rr->br.y+1, rr->br.x,rr->br.y+1, "+"); } -static void find_table_entry(Rect commod, int colno, Rect *cell) { +static void compute_table_location(Rect commod, int colno, Rect *cell) { cell->tl.y= commod.tl.y; cell->br.y= commod.br.y; - cell->tl.x= !colno ? commod.tl.x : colrightx[colno-1]+2; - cell->br.x= colrightx[colno]; + cell->tl.x= !colno ? commod.tl.x : s.colrightx[colno-1]+2; + cell->br.x= s.colrightx[colno]; debug_rect("cell", colno, *cell); +} - const RgbImage *ri= cim->rgb; - +static void ocr_rectangle(Rect r, const OcrCellType ct, FILE *tsv_output) { + OcrResultGlyph *results, *res; + + int w= r.br.x - r.tl.x + 1; + Pixcol cols[w+1]; + int x,y; + for (x=0; x= '0' && pixel <= '0'+AAMAXVAL, + MC(pixel);MP(here);MSB(ocr_celltype_name(ct));MR(r); ); + pixcol_p_add(&cols[x], y, pixel-'0'); + } + } + FILLZERO(cols[w]); + + results= ocr(rd,ct,w,cols); + for (res=results; res->s; res++) + fputs(res->s,tsv_output); +} + +#define FOR_COMMODITY_CELL(ROW_START, CELL, ROW_END) do{ \ + Rect rowr, cell; \ + int tryrect, colno; \ + \ + for (tryrect= +cim->h; tryrect >= -cim->h; tryrect--) { \ + find_commodity(tryrect, &rowr); \ + if (rowr.tl.x < 0) \ + continue; \ + debug_rect("commod",tryrect, rowr); \ + \ + ROW_START; \ + \ + for (colno=0; colnobr.x, cell->br.y); - memcpy(chanbg, RI_PIXEL(ri, cell->br.x, cell->br.y), 3); - - FOR_P_RECT(p,*cell) { - const unsigned char *here_pixel= RI_PIXEL(ri, p.x, p.y); - int i; - for (i=0; i<3; i++) { - unsigned here= here_pixel[i]; - if (here == chanbg[i]) bg_count++; - else if (here < chanbg[i]) dark_count += (chanbg[i] - here)/4 + 1; - else if (here > chanbg[i]) light_count += (here - chanbg[i])/4 + 1; + background= ri_rgb(ri, cell.br.x, cell.br.y); + for (i=0; i<3; i++) + chanbg[i]= background >> (i*8); + + FOR_P_RECT(p,cell) { + Rgb herergb= ri_rgb(ri, p.x, p.y); + if (herergb==background) { + bg_count+=3; + } else { + for (i=0; i<3; i++) { + unsigned char here= herergb >> (i*8); + if (here == chanbg[i]) bg_count++; + else if (here < chanbg[i]) dark_count += (chanbg[i] - here)/4 + 1; + else if (here > chanbg[i]) light_count += (here - chanbg[i])/4 + 1; + } } } - long total_count= RECT_W(*cell) * RECT_H(*cell) * 3; + long total_count= RECT_W(cell) * RECT_H(cell) * 3; MUST( bg_count > total_count / 2, - MR(*cell);MIL(total_count);MIL(bg_count); + MR(cell);MIL(total_count);MIL(bg_count); MIL(light_count);MIL(dark_count) ); if (bg_count == total_count) @@ -615,60 +745,48 @@ static void find_table_entry(Rect commod, int colno, Rect *cell) { fg_extra= -1; } else { MUST( !"tell light from dark", - MR(*cell);MIL(total_count);MIL(bg_count); + MR(cell);MIL(total_count);MIL(bg_count); MIL(light_count);MIL(dark_count);MRGB(background); ); } debugf("TABLEENTRY col=%d %d,%d..%d,%d bg=%ld light=%ld dark=%ld\n", - colno, cell->tl.x,cell->tl.y, cell->br.x,cell->br.y, + colno, cell.tl.x,cell.tl.y, cell.br.x,cell.br.y, bg_count, light_count, dark_count); - + int monochrome= 1; - FOR_P_RECT(p,*cell) { - double alpha= find_aa_density(ri,p,background,foreground,fg_extra); + find_aa_density_prep(background, foreground, fg_extra); - int here_int= floor((AAMAXVAL+1)*alpha); + FOR_P_RECT(p,cell) { + Fixpt alpha= find_aa_density(ri,p); + + int here_int= alpha >> (FIXPT_SHIFT - AADEPTH); assert(here_int <= AAMAXVAL); if (!(here_int==0 || here_int==AAMAXVAL)) monochrome=0; - cim->d[p.y * cim->w + p.x]= '0' + here_int; + ci->d[p.y * ci->w + p.x]= '0' + here_int; } - debug_rect("cell0M", colno, *cell); + debug_rect("cell0M", colno, cell); - require_rectangle_r(*cell, "0123456789", __LINE__); + require_rectangle_r(cell, "0123456789", __LINE__); } -static void ocr_rectangle(Rect r, const OcrCellType ct, FILE *tsv_output) { - OcrResultGlyph *results, *res; +void adjust_colours(CanonImage *ci, const RgbImage *ri) { + if (!(o_mode & mf_analyse)) + return; - int w= r.br.x - r.tl.x + 1; - Pixcol cols[w+1]; - int x,y; - for (x=0; x= '0' && pixel <= '0'+AAMAXVAL, - MC(pixel);MP(here);MSB(ocr_celltype_name(ct));MR(r); ); - pixcol_p_add(&cols[x], y, pixel-'0'); - } - } - FILLZERO(cols[w]); + cim= ci; - results= ocr(rd,ct,w,cols); - for (res=results; res->s; res++) - fputs(res->s,tsv_output); + FOR_COMMODITY_CELL({},({ + adjust_colours_cell(ci,ri,colno,cell); + }),{}); } void analyse(FILE *tsv_output) { - Rect thisr, entryr; - int page, tryrect, colno; + int page; for (page=0; pageh; 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; colnorgb->w, - page_images[0]->rgb->h); - memcpy(ri->data, page_images[0]->rgb->data, ri->w * ri->h * 3); + RgbImage *ri= alloc_rgb_image(rgbsrc->w, rgbsrc->h); + memcpy(ri->data, rgbsrc->data, ri->w * ri->h * 3); Rect sunshiner= find_sunshine_widget(); char sunshine[MAXIMGIDENT], archisland[MAXIMGIDENT]; - const unsigned char *srcp; - unsigned char *destp, *endp; - for (srcp=page_images[0]->rgb->data, destp=ri->data, - endp= ri->data + 3 * ri->w * ri->h; + const Rgb *srcp; + Rgb *destp, *endp; + for (srcp= rgbsrc->data, destp=ri->data, + endp= ri->data + ri->w * ri->h; destp < endp; srcp++, destp++) { - unsigned char c= *srcp & 0xf0; - *destp= c | (c>>4); + Rgb new= *srcp & 0xf0f0f0; + *destp= new | (new>>4); } identify_rgbimage(ri, sunshiner, sunshine, "sunshine widget"); @@ -771,13 +885,11 @@ void find_islandname(void) { // islandnamer.br.y = larger_islandnamebry; debug_rect("islandnamer",1, islandnamer); - int x,y,i; + int x,y; for (x=islandnamer.tl.x; x<=islandnamer.br.x; x++) for (y=islandnamer.tl.y; y<=islandnamer.br.y; y++) { - if (RI_PIXEL(ri,x,y)[0] < 0x40) { - for (i=0; i<3; i++) { - RI_PIXEL(ri,x,y)[i]= 0; - } + if ((ri_rgb(ri,x,y) & 0xff) < 0x40) { + *RI_PIXEL32(ri,x,y)= 0; } } @@ -830,11 +942,13 @@ void find_islandname(void) { uint32_t pattern=0; int runs[32], nruns=0; runs[0]=0; runs[1]=0; + + find_aa_density_prep(0xCCCCAA,0x002255,0); for (p.y=islandnamer.tl.y; p.y<=islandnamer.br.y; p.y++) { pattern <<= 1; - double alpha= find_aa_density(ri,p, 0xCCCCAA,0x002255,0); - if (alpha >= 0.49) { + Fixpt alpha= find_aa_density(ri,p); + if (alpha >= dbl2fixpt(0.49)) { runs[nruns]++; pattern |= 1u; } else { diff --git a/pctb/structure.h b/pctb/structure.h index 84fdcae..263a26a 100644 --- a/pctb/structure.h +++ b/pctb/structure.h @@ -50,44 +50,41 @@ static inline char canon_lookup_colour(unsigned char r, return blues->blue2[b]; } -#define CANONICALISE_IMAGE(im,w,h, COMPUTE_RGB) do{ \ - /* compute_rgb should be a number of statements, or \ - * a block, which assigns to \ - * Rgb rgb; \ - * given the values of \ - * int x,y; \ - * all of which are anamorphic. Result is stored in im. \ - * The COMPUTE_RGB is executed exactly once for \ - * each pixel in reading order. \ - */ \ - (im)= alloc_canon_image((w), (h)); \ - RgbImage *rgb_save; \ - (im)->rgb= rgb_save= alloc_rgb_image((w), (h)); \ - \ - int x,y; \ - for (y=0; y<(h); y++) { \ - for (x=0; x<(w); x++) { \ - unsigned char r,g,b; \ - COMPUTE_RGB; \ - CANONIMG_ALSO_STORERGB(rgb_save); \ - (im)->d[y*(w) + x]= canon_lookup_colour(r,g,b); \ - } \ - if (DEBUGP(rect)) { \ - fprintf(debug, "%4d ",y); \ - fwrite(im->d + y*(w), 1,(w), debug); \ - fputc('\n',debug); \ - } \ - } \ - debug_flush(); \ +#define CANONICALISE_IMAGE(im,w,h,rgb_save, COMPUTE_RGB) do{ \ + /* compute_rgb should be a number of statements, or \ + * a block, which assigns to \ + * Rgb rgb; \ + * given the values of \ + * int x,y; \ + * all of which are anamorphic. Result is stored in im. \ + * The COMPUTE_RGB is executed exactly once for \ + * each pixel in reading order. \ + */ \ + (im)= alloc_canon_image((w), (h)); \ + (rgb_save)= alloc_rgb_image((w), (h)); \ + \ + int x,y; \ + for (y=0; y<(h); y++) { \ + for (x=0; x<(w); x++) { \ + Rgb rgb; \ + COMPUTE_RGB; \ + CANONIMG_ALSO_STORERGB((rgb_save)); \ + (im)->d[y*(w) + x]= canon_lookup_colour(rgb, rgb>>8, rgb>>16); \ + } \ + if (DEBUGP(rect)) { \ + fprintf(debug, "%4d ",y); \ + fwrite(im->d + y*(w), 1,(w), debug); \ + fputc('\n',debug); \ + } \ + } \ + debug_flush(); \ }while(0) #define CANONIMG_ALSO_STORERGB(ri) \ do{ \ - unsigned char *rip= RI_PIXEL((ri),x,y); \ - rip[0]= r; \ - rip[1]= g; \ - rip[2]= b; \ + Rgb *rip= RI_PIXEL32((ri),x,y); \ + *rip= rgb; \ }while(0) diff --git a/pctb/x.gdb b/pctb/x.gdb index 0f6bf86..1133843 100644 --- a/pctb/x.gdb +++ b/pctb/x.gdb @@ -1,4 +1,5 @@ file ypp-commodities -set args -Drect 2>u --edit-charset --raw-tsv --same --screenshot-file t.ppm >raw.tsv +set args -Drect 2>u --edit-charset --find-island --ocean sage +#break structure.c:596 if here!=aa_background break mustfail2 run