/* Copyright (C) 2001-2006 Artifex Software, Inc. All Rights Reserved. This software is provided AS-IS with no warranty, either express or implied. This software is distributed under license and may not be copied, modified or distributed except as expressly authorized under the terms of that license. Refer to licensing information at http://www.artifex.com/ or contact Artifex Software, Inc., 7 Mt. Lassen Drive - Suite A-134, San Rafael, CA 94903, U.S.A., +1(415)492-9861, for further information. */ /* $Id$ */ /* PDF 1.4 blending functions */ #include "memory_.h" #include "gx.h" #include "gstparam.h" #include "gsrect.h" #include "gxblend.h" #include "gxdcconv.h" #include "gxdevcli.h" #include "gxistate.h" #include "gdevdevn.h" #include "gdevp14.h" #include "vdtrace.h" #include "gxdcconv.h" #ifdef DUMP_TO_PNG #include "png_.h" #endif /* * Unpack a device color. This routine is similar to the device's * decode_color procedure except for two things. The procedure produces 1 * byte values instead of gx_color_values (2 bytes). A separate * procedure is used instead of the decode_color to minimize execution time. */ void pdf14_unpack_additive(int num_comp, gx_color_index color, pdf14_device * p14dev, byte * out) { int i; for (i = num_comp - 1; i >= 0; i--) { out[i] = (byte)(color & 0xff); color >>= 8; } } /* * Unpack a device color. This routine is similar to the device's * decode_color procedure except for two things. The procedure produces 1 * byte values instead of gx_color_values (2 bytes) and the output values * are inverted for subtractive color spaces (like CMYK). A separate * procedure is used instead of the decode_color to minimize execution time. */ void pdf14_unpack_subtractive(int num_comp, gx_color_index color, pdf14_device * p14dev, byte * out) { int i; for (i = num_comp - 1; i >= 0; i--) { out[i] = 0xff - (byte)(color & 0xff); color >>= 8; } } /* * Unpack a 'compressed' CMYK color index. The color index value is unpacked * into a set of 8 bit values. For more information about 'compressed' color * index values see the comments before the devn_encode_compressed_color routine. * * Note: For simplicity of coding the calling routines, this routine will also * handle 'uncompressed' color index values. */ void pdf14_unpack_compressed(int num_comp, gx_color_index color, pdf14_device * p14dev, byte * out) { int comp_num; if (p14dev->devn_params.compressed_color_list == NULL) { /* * For 'uncompressed' data we simply have to unpack the gx_color_index * value directly. */ for (comp_num = num_comp - 1; comp_num >= 0; comp_num--) { out[comp_num] = 0xff - (byte)(color & 0xff); color >>= 8; } } else { int factor, bit_count, bit_mask; comp_bit_map_list_t * pbitmap; gx_color_value solid_color = 0xff; pbitmap = find_bit_map(color, p14dev->devn_params.compressed_color_list); bit_count = num_comp_bits[pbitmap->num_non_solid_comp]; bit_mask = (1 << bit_count) - 1; factor = comp_bit_factor[pbitmap->num_non_solid_comp]; if (pbitmap->solid_not_100) { solid_color = 0xff - ((factor * ((int)color & bit_mask)) >> 16); color >>= bit_count; } for (comp_num = 0; comp_num < num_comp; comp_num++) { if (colorant_present(pbitmap, colorants, comp_num)) { if (colorant_present(pbitmap, solid_colorants, comp_num)) *out++ = (byte)solid_color; else { *out++ = 0xff - ((factor * ((int)color & bit_mask)) >> 16); color >>= bit_count; } } else *out++ = 0xff; } } } /* * Unpack a device color. This routine is used for devices in which we do * not know the details of the process color model. In this case we use * the device's decode_color procedure. */ void pdf14_unpack_custom(int num_comp, gx_color_index color, pdf14_device * p14dev, byte * out) { int i; gx_device * tdev = p14dev->target; gx_color_value cm_values[GX_DEVICE_COLOR_MAX_COMPONENTS]; dev_proc(tdev, decode_color)(tdev, color, cm_values); for (i = 0; i < num_comp; i++) out[i] = 0xff - gx_color_value_to_byte(cm_values[i]); } #if RAW_DUMP extern unsigned int global_index; #endif void pdf14_preserve_backdrop(pdf14_buf *buf, pdf14_buf *tos, bool has_shape) { /* make copy of backdrop for compositing */ int x0 = max(buf->rect.p.x, tos->rect.p.x); int x1 = min(buf->rect.q.x, tos->rect.q.x); int y0 = max(buf->rect.p.y, tos->rect.p.y); int y1 = min(buf->rect.q.y, tos->rect.q.y); if (x0 < x1 && y0 < y1) { int width = x1 - x0; byte *buf_plane = buf->data + x0 - buf->rect.p.x + (y0 - buf->rect.p.y) * buf->rowstride; byte *tos_plane = tos->data + x0 - tos->rect.p.x + (y0 - tos->rect.p.y) * tos->rowstride; int i; /*int n_chan_copy = buf->n_chan + (tos->has_shape ? 1 : 0);*/ int n_chan_copy = tos->n_chan + (tos->has_shape ? 1 : 0) + (tos->has_tags ? 1 : 0); for (i = 0; i < n_chan_copy; i++) { byte *buf_ptr = buf_plane; byte *tos_ptr = tos_plane; int y; for (y = y0; y < y1; ++y) { memcpy (buf_ptr, tos_ptr, width); buf_ptr += buf->rowstride; tos_ptr += tos->rowstride; } buf_plane += buf->planestride; tos_plane += tos->planestride; } if (has_shape && !tos->has_shape) { if (tos->has_tags) { buf_plane -= buf->planestride; } memset (buf_plane, 0, buf->planestride); } } #if RAW_DUMP if (x0 < x1 && y0 < y1) { byte *buf_plane = buf->data + x0 - buf->rect.p.x + (y0 - buf->rect.p.y) * buf->rowstride; dump_raw_buffer(y1-y0, x1 - x0, buf->n_planes, buf->planestride, buf->rowstride, "BackDropInit",buf_plane); global_index++; } #endif } void pdf14_compose_group(pdf14_buf *tos, pdf14_buf *nos, pdf14_buf *maskbuf, int x0, int x1, int y0, int y1, int n_chan, bool additive, const pdf14_nonseparable_blending_procs_t * pblend_procs) { int num_comp = n_chan - 1; byte alpha = tos->alpha; byte shape = tos->shape; byte blend_mode = tos->blend_mode; byte *tos_ptr = tos->data + x0 - tos->rect.p.x + (y0 - tos->rect.p.y) * tos->rowstride; byte *nos_ptr = nos->data + x0 - nos->rect.p.x + (y0 - nos->rect.p.y) * nos->rowstride; byte *mask_ptr = NULL; int tos_planestride = tos->planestride; int nos_planestride = nos->planestride; int mask_planestride = 0x0badf00d; /* Quiet compiler. */ byte mask_bg_alpha = 0; /* Quiet compiler. */ int width = x1 - x0; int x, y; int i, tmp; byte tos_pixel[PDF14_MAX_PLANES]; byte nos_pixel[PDF14_MAX_PLANES]; bool tos_isolated = tos->isolated; bool nos_knockout = nos->knockout; byte *nos_alpha_g_ptr; int tos_shape_offset = n_chan * tos_planestride; int tos_alpha_g_offset = tos_shape_offset + (tos->has_shape ? tos_planestride : 0); bool tos_has_tag = tos->has_tags; int tos_tag_offset = tos_planestride * (tos->n_planes - 1); int nos_shape_offset = n_chan * nos_planestride; bool nos_has_shape = nos->has_shape; bool nos_has_tag = nos->has_tags; int nos_tag_offset = nos_planestride * (nos->n_planes - 1); byte *mask_tr_fn = NULL; /* Quiet compiler. */ #if RAW_DUMP byte *composed_ptr = NULL; #endif if ((tos->n_chan == 0) || (nos->n_chan == 0)) return; rect_merge(nos->dirty, tos->dirty); if_debug6('v', "pdf14_pop_transparency_group y0 = %d, y1 = %d, w = %d, alpha = %d, shape = %d, tag = bm = %d\n", y0, y1, width, alpha, shape, blend_mode); if (nos->has_alpha_g) nos_alpha_g_ptr = nos_ptr + n_chan * nos_planestride; else nos_alpha_g_ptr = NULL; if (maskbuf != NULL) { mask_tr_fn = maskbuf->transfer_fn; if (maskbuf->data != NULL) { mask_ptr = maskbuf->data + x0 - maskbuf->rect.p.x + (y0 - maskbuf->rect.p.y) * maskbuf->rowstride; mask_planestride = maskbuf->planestride; } else { /* We may have a case, where we are outside the maskbuf rect. */ /* We would have avoided creating the maskbuf->data */ /* In that case, we should use the background alpha value */ /* See discussion on the BC entry in the PDF spec */ mask_bg_alpha = maskbuf->alpha; /* Adjust alpha by the mask background alpha */ mask_bg_alpha = mask_tr_fn[mask_bg_alpha]; tmp = alpha * mask_bg_alpha + 0x80; alpha = (tmp + (tmp >> 8)) >> 8; } } #if RAW_DUMP composed_ptr = nos_ptr; dump_raw_buffer(y1-y0, width, tos->n_planes, tos_planestride, tos->rowstride, "ImageTOS",tos_ptr); dump_raw_buffer(y1-y0, width, nos->n_planes, nos_planestride, nos->rowstride, "ImageNOS",nos_ptr); if(mask_ptr != NULL){ dump_raw_buffer(y1-y0, width, maskbuf->n_planes, maskbuf->planestride, maskbuf->rowstride, "Mask",mask_ptr); } #endif for (y = y0; y < y1; ++y) { for (x = 0; x < width; ++x) { byte pix_alpha = alpha; /* Complement the components for subtractive color spaces */ if (additive) { for (i = 0; i < n_chan; ++i) { tos_pixel[i] = tos_ptr[x + i * tos_planestride]; nos_pixel[i] = nos_ptr[x + i * nos_planestride]; } } else { for (i = 0; i < num_comp; ++i) { tos_pixel[i] = 255 - tos_ptr[x + i * tos_planestride]; nos_pixel[i] = 255 - nos_ptr[x + i * nos_planestride]; } tos_pixel[num_comp] = tos_ptr[x + num_comp * tos_planestride]; nos_pixel[num_comp] = nos_ptr[x + num_comp * nos_planestride]; } if (mask_ptr != NULL) { byte mask = mask_ptr[x]; mask = mask_tr_fn[mask]; tmp = pix_alpha * mask + 0x80; pix_alpha = (tmp + (tmp >> 8)) >> 8; # if VD_PAINT_MASK vd_pixel(int2fixed(x), int2fixed(y), mask); # endif } if (nos_knockout) { byte *nos_shape_ptr = nos_has_shape ? &nos_ptr[x + nos_shape_offset] : NULL; byte *nos_tag_ptr = nos_has_tag ? &nos_ptr[x + nos_tag_offset] : NULL; byte tos_shape = tos_ptr[x + tos_shape_offset]; byte tos_tag = tos_ptr[x + tos_tag_offset]; art_pdf_composite_knockout_isolated_8(nos_pixel, nos_shape_ptr, nos_tag_ptr, tos_pixel, n_chan - 1, tos_shape, tos_tag, pix_alpha, shape); } else { if (tos_isolated) { art_pdf_composite_group_8(nos_pixel, nos_alpha_g_ptr, tos_pixel, n_chan - 1, pix_alpha, blend_mode, pblend_procs); } else { byte tos_alpha_g = tos_ptr[x + tos_alpha_g_offset]; art_pdf_recomposite_group_8(nos_pixel, nos_alpha_g_ptr, tos_pixel, tos_alpha_g, n_chan - 1, pix_alpha, blend_mode, pblend_procs); } if (tos_has_tag) { if (pix_alpha == 255) { nos_ptr[x + nos_tag_offset] = tos_ptr[x + tos_tag_offset]; } else if (pix_alpha != 0 && tos_ptr[x + tos_tag_offset] != GS_UNTOUCHED_TAG) { nos_ptr[x + nos_tag_offset] = (nos_ptr[x + nos_tag_offset] | tos_ptr[x + tos_tag_offset]) & ~GS_UNTOUCHED_TAG; } } } if (nos_has_shape) { nos_ptr[x + nos_shape_offset] = art_pdf_union_mul_8 (nos_ptr[x + nos_shape_offset], tos_ptr[x + tos_shape_offset], shape); } /* Complement the results for subtractive color spaces */ if (additive) { for (i = 0; i < n_chan; ++i) { nos_ptr[x + i * nos_planestride] = nos_pixel[i]; } } else { for (i = 0; i < num_comp; ++i) nos_ptr[x + i * nos_planestride] = 255 - nos_pixel[i]; nos_ptr[x + num_comp * nos_planestride] = nos_pixel[num_comp]; } # if VD_PAINT_COLORS vd_pixel(int2fixed(x), int2fixed(y), n_chan == 1 ? (nos_pixel[0] << 16) + (nos_pixel[0] << 8) + nos_pixel[0] : (nos_pixel[0] << 16) + (nos_pixel[1] << 8) + nos_pixel[2]); # endif # if VD_PAINT_ALPHA vd_pixel(int2fixed(x), int2fixed(y), (nos_pixel[n_chan - 1] << 16) + (nos_pixel[n_chan - 1] << 8) + nos_pixel[n_chan - 1]); # endif if (nos_alpha_g_ptr != NULL) ++nos_alpha_g_ptr; } tos_ptr += tos->rowstride; nos_ptr += nos->rowstride; if (nos_alpha_g_ptr != NULL) nos_alpha_g_ptr += nos->rowstride - width; if (mask_ptr != NULL) mask_ptr += maskbuf->rowstride; } /* Lets look at composed result */ #if RAW_DUMP /* The group alpha should disappear */ dump_raw_buffer(y1-y0, width, tos->n_planes - tos->has_alpha_g - tos->has_shape, nos_planestride, nos->rowstride, "Composed",composed_ptr); global_index++; #endif } /* * Encode a list of colorant values into a gx_color_index_value. */ gx_color_index pdf14_encode_color(gx_device *dev, const gx_color_value colors[]) { gx_color_index color = 0; int i; int ncomp = dev->color_info.num_components; COLROUND_VARS; COLROUND_SETUP(8); for (i = 0; i < ncomp; i++) { color <<= 8; color |= COLROUND_ROUND(colors[i]); } return (color == gx_no_color_index ? color ^ 1 : color); } /* * Encode a list of colorant values into a gx_color_index_value. Stick the tag information at the end. */ gx_color_index pdf14_encode_color_tag(gx_device *dev, const gx_color_value colors[]) { gx_color_index color; int i; int ncomp = dev->color_info.num_components; COLROUND_VARS; COLROUND_SETUP(8); /* Add in the tag information */ color = dev->graphics_type_tag & ~GS_DEVICE_ENCODES_TAGS; for (i = 0; i < ncomp; i++) { color <<= 8; color |= COLROUND_ROUND(colors[i]); } return (color == gx_no_color_index ? color ^ 1 : color); } /* * Decode a gx_color_index value back to a list of colorant values. */ int pdf14_decode_color(gx_device * dev, gx_color_index color, gx_color_value * out) { int i; int ncomp = dev->color_info.num_components; for (i = 0; i < ncomp; i++) { out[ncomp - i - 1] = (gx_color_value) ((color & 0xff) * 0x101); color >>= 8; } return 0; } /* * Encode a list of colorant values into a gx_color_index_value. For more * information about 'compressed' color index values see the comments before * the devn_encode_compressed_color routine. */ gx_color_index pdf14_compressed_encode_color(gx_device *dev, const gx_color_value colors[]) { gs_devn_params *pdevn_params = NULL; if (dev->procs.ret_devn_params != NULL) pdevn_params = dev_proc(dev, ret_devn_params)(dev); /* If there was no dev_proc or it returned NULL, assume pdf14 device devn_params */ if (pdevn_params == NULL) { #ifdef DEBUG if (strncmp(dev->dname, "pdf14", 5)) emprintf1(dev->memory, "pdf14_compressed_encode_color devn_params not from pdf14 device, device = '%s'\n", dev->dname); #endif pdevn_params = &(((pdf14_device *)dev)->devn_params); } return devn_encode_compressed_color(dev, colors, pdevn_params); } /* * Decode a gx_color_index value back to a list of colorant values. For more * information about 'compressed' color index values see the comments before * the devn_encode_compressed_color routine. */ int pdf14_compressed_decode_color(gx_device * dev, gx_color_index color, gx_color_value * out) { gs_devn_params *pdevn_params = NULL; if (dev->procs.ret_devn_params != NULL) pdevn_params = dev_proc(dev, ret_devn_params)(dev); /* If there was no dev_proc or it returned NULL, assume pdf14 device devn_params */ if (pdevn_params == NULL) { #ifdef DEBUG if (strncmp(dev->dname, "pdf14", 5)) emprintf1(dev->memory, "pdf14_compressed_decode_color devn_params not from pdf14 device, device = '%s'\n", dev->dname); #endif pdevn_params = &(((pdf14_device *)dev)->devn_params); } return devn_decode_compressed_color(dev, color, out, pdevn_params); } void pdf14_gray_cs_to_cmyk_cm(gx_device * dev, frac gray, frac out[]) { int num_comp = dev->color_info.num_components; out[0] = out[1] = out[2] = frac_0; out[3] = frac_1 - gray; for (--num_comp; num_comp > 3; num_comp--) out[num_comp] = 0; } /* * Default map from DeviceRGB color space to DeviceCMYK color * model. Since this mapping is defined by the PostScript language * it is unlikely that any device with a DeviceCMYK color model * would define this mapping on its own. * * If the imager state is not available, map as though the black * generation and undercolor removal functions are identity * transformations. This mode is used primarily to support the * raster operation (rop) feature of PCL, which requires that * the raster operation be performed in an RGB color space. * Note that default black generation and undercolor removal * functions in PostScript need NOT be identity transformations: * often they are { pop 0 }. */ void pdf14_rgb_cs_to_cmyk_cm(gx_device * dev, const gs_imager_state *pis, frac r, frac g, frac b, frac out[]) { int num_comp = dev->color_info.num_components; if (pis != 0) color_rgb_to_cmyk(r, g, b, pis, out, dev->memory); else { frac c = frac_1 - r, m = frac_1 - g, y = frac_1 - b; frac k = min(c, min(m, g)); out[0] = c - k; out[1] = m - k; out[2] = y - k; out[3] = k; } for (--num_comp; num_comp > 3; num_comp--) out[num_comp] = 0; } void pdf14_cmyk_cs_to_cmyk_cm(gx_device * dev, frac c, frac m, frac y, frac k, frac out[]) { int num_comp = dev->color_info.num_components; out[0] = c; out[1] = m; out[2] = y; out[3] = k; for (--num_comp; num_comp > 3; num_comp--) out[num_comp] = 0; } #ifdef DUMP_TO_PNG /* Dumps a planar RGBA image to a PNG file. */ static int dump_planar_rgba(gs_memory_t *mem, const pdf14_buf *pbuf) { int rowstride = pbuf->rowstride, planestride = pbuf->planestride; int rowbytes = width << 2; gs_int_rect rect = buf->rect; int x1 = min(pdev->width, rect.q.x); int y1 = min(pdev->height, rect.q.y); int width = x1 - rect.p.x; int height = y1 - rect.p.y; byte *buf_ptr = buf->data + rect.p.y * buf->rowstride + rect.p.x; byte *row = gs_malloc(mem, rowbytes, 1, "png raster buffer"); png_struct *png_ptr = png_create_write_struct(PNG_LIBPNG_VER_STRING, NULL, NULL, NULL); png_info *info_ptr = png_create_info_struct(png_ptr); const char *software_key = "Software"; char software_text[256]; png_text text_png; FILE *file; int code; int y; if (buf->data == NULL) return 0; file = fopen ("c:\\temp\\tmp.png", "wb"); if_debug0('v', "[v]pnga_output_page\n"); if (row == 0 || png_ptr == 0 || info_ptr == 0) { code = gs_note_error(gs_error_VMerror); goto done; } /* set error handling */ if (setjmp(png_ptr->jmpbuf)) { /* If we get here, we had a problem reading the file */ code = gs_note_error(gs_error_VMerror); goto done; } code = 0; /* for normal path */ /* set up the output control */ png_init_io(png_ptr, file); /* set the file information here */ info_ptr->width = width; info_ptr->height = height; /* resolution is in pixels per meter vs. dpi */ info_ptr->x_pixels_per_unit = (png_uint_32) (96.0 * (100.0 / 2.54)); info_ptr->y_pixels_per_unit = (png_uint_32) (96.0 * (100.0 / 2.54)); info_ptr->phys_unit_type = PNG_RESOLUTION_METER; info_ptr->valid |= PNG_INFO_pHYs; /* At present, only supporting 32-bit rgba */ info_ptr->bit_depth = 8; info_ptr->color_type = PNG_COLOR_TYPE_RGB_ALPHA; /* add comment */ sprintf(software_text, "%s %d.%02d", gs_product, (int)(gs_revision / 100), (int)(gs_revision % 100)); text_png.compression = -1; /* uncompressed */ text_png.key = (char *)software_key; /* not const, unfortunately */ text_png.text = software_text; text_png.text_length = strlen(software_text); info_ptr->text = &text_png; info_ptr->num_text = 1; /* write the file information */ png_write_info(png_ptr, info_ptr); /* don't write the comments twice */ info_ptr->num_text = 0; info_ptr->text = NULL; /* Write the contents of the image. */ for (y = 0; y < height; ++y) { int x; for (x = 0; x < width; ++x) { row[(x << 2)] = buf_ptr[x]; row[(x << 2) + 1] = buf_ptr[x + planestride]; row[(x << 2) + 2] = buf_ptr[x + planestride * 2]; row[(x << 2) + 3] = buf_ptr[x + planestride * 3]; } png_write_row(png_ptr, row); buf_ptr += rowstride; } /* write the rest of the file */ png_write_end(png_ptr, info_ptr); done: /* free the structures */ png_destroy_write_struct(&png_ptr, &info_ptr); gs_free(mem, row, rowbytes, 1, "png raster buffer"); fclose (file); return code; } #endif void gx_build_blended_image_row(byte *buf_ptr, int y, int planestride, int width, int num_comp, byte bg, byte *linebuf) { int x; for (x = 0; x < width; x++) { byte comp, a; int tmp, comp_num; /* composite RGBA (or CMYKA, etc.) pixel with over solid background */ a = buf_ptr[x + planestride * num_comp]; if ((a + 1) & 0xfe) { a ^= 0xff; for (comp_num = 0; comp_num < num_comp; comp_num++) { comp = buf_ptr[x + planestride * comp_num]; tmp = ((bg - comp) * a) + 0x80; comp += (tmp + (tmp >> 8)) >> 8; linebuf[x * num_comp + comp_num] = comp; } } else if (a == 0) { for (comp_num = 0; comp_num < num_comp; comp_num++) { linebuf[x * num_comp + comp_num] = bg; } } else { for (comp_num = 0; comp_num < num_comp; comp_num++) { comp = buf_ptr[x + planestride * comp_num]; linebuf[x * num_comp + comp_num] = comp; } } } } void gx_blend_image_buffer(byte *buf_ptr, int width, int height, int rowstride, int planestride, int num_comp, byte bg) { int x, y; int position; byte comp, a; int tmp, comp_num; for (y = 0; y < height; y++) { position = y * rowstride; for (x = 0; x < width; x++) { /* composite RGBA (or CMYKA, etc.) pixel with over solid background */ a = buf_ptr[position + planestride * num_comp]; if ((a + 1) & 0xfe) { a ^= 0xff; for (comp_num = 0; comp_num < num_comp; comp_num++) { comp = buf_ptr[position + planestride * comp_num]; tmp = ((bg - comp) * a) + 0x80; comp += (tmp + (tmp >> 8)) >> 8; buf_ptr[position + planestride * comp_num] = comp; } } else if (a == 0) { for (comp_num = 0; comp_num < num_comp; comp_num++) { buf_ptr[position + planestride * comp_num] = bg; } } position+=1; } } } int gx_put_blended_image_cmykspot(gx_device *target, byte *buf_ptr, int planestride, int rowstride, int x0, int y0, int width, int height, int num_comp, byte bg, gs_separations * pseparations) { int code = 0; int x, y, tmp, comp_num, output_comp_num; gx_color_index color; gx_color_value cv[GX_DEVICE_COLOR_MAX_COMPONENTS]; gx_color_value comp; byte a; int input_map[GX_DEVICE_COLOR_MAX_COMPONENTS]; int output_map[GX_DEVICE_COLOR_MAX_COMPONENTS]; int num_known_comp = 0; int output_num_comp = target->color_info.num_components; int num_sep = pseparations->num_separations++; /* * The process color model for the PDF 1.4 compositor device is CMYK plus * spot colors. The target device may have only some of these colorants due * to the SeparationOrder device parameter. Thus we need to determine the * mapping between the PDF 1.4 compositor and the target device. Note: * There should not be a spot colorant in the PDF 1.4 device that is not * present in the target device. */ /* Check if target processes CMYK colorants. */ for (comp_num = 0; comp_num < 4; comp_num++) { const char * pcomp_name = (const char *)DeviceCMYKComponents[comp_num]; output_comp_num = dev_proc(target, get_color_comp_index) (target, pcomp_name, strlen(pcomp_name), NO_COMP_NAME_TYPE); if (output_comp_num >=0 && output_comp_num < GX_DEVICE_COLOR_MAX_COMPONENTS) { output_map[num_known_comp] = output_comp_num; input_map[num_known_comp++] = comp_num; } } /* Check if target processes our spot colorants. */ for (comp_num = 0; comp_num < num_sep; comp_num++) { output_comp_num = dev_proc(target, get_color_comp_index) (target, (const char *)(pseparations->names[comp_num].data), pseparations->names[comp_num].size, NO_COMP_NAME_TYPE); if (output_comp_num >= 0 && output_comp_num < GX_DEVICE_COLOR_MAX_COMPONENTS) { output_map[num_known_comp] = output_comp_num; input_map[num_known_comp++] = comp_num + 4; } } /* Clear all output colorants first */ for (comp_num = 0; comp_num < output_num_comp; comp_num++) cv[comp_num] = 0; /* Send pixel data to the target device. */ for (y = 0; y < height; y++) { for (x = 0; x < width; x++) { /* composite CMYKA, etc. pixel with over solid background */ a = buf_ptr[x + planestride * num_comp]; if ((a + 1) & 0xfe) { /* a ^= 0xff; */ /* No inversion here! Bug 689895 */ for (comp_num = 0; comp_num < num_known_comp; comp_num++) { comp = buf_ptr[x + planestride * input_map[comp_num]]; tmp = ((comp - bg) * a) + 0x80; comp += tmp + (tmp >> 8); cv[output_map[comp_num]] = comp; } } else if (a == 0) { for (comp_num = 0; comp_num < num_known_comp; comp_num++) { cv[output_map[comp_num]] = bg; } } else { for (comp_num = 0; comp_num < num_known_comp; comp_num++) { comp = buf_ptr[x + planestride * input_map[comp_num]]; cv[output_map[comp_num]] = (comp << 8) + comp; } } color = dev_proc(target, encode_color)(target, cv); code = dev_proc(target, fill_rectangle)(target, x + x0, y + y0, 1, 1, color); if (code < 0) return code; } buf_ptr += rowstride; } return code; } int gx_put_blended_image_custom(gx_device *target, byte *buf_ptr, int planestride, int rowstride, int x0, int y0, int width, int height, int num_comp, byte bg) { int code = 0; int x, y, tmp, comp_num; gx_color_index color; gx_color_value cv[GX_DEVICE_COLOR_MAX_COMPONENTS]; gx_color_value comp; byte a; /* Send pixel data to the target device. */ for (y = 0; y < height; y++) { for (x = 0; x < width; x++) { /* composite CMYKA, etc. pixel with over solid background */ a = buf_ptr[x + planestride * num_comp]; if ((a + 1) & 0xfe) { a ^= 0xff; for (comp_num = 0; comp_num < num_comp; comp_num++) { comp = buf_ptr[x + planestride * comp_num]; tmp = ((bg - comp) * a) + 0x80; comp += tmp + (tmp >> 8); cv[comp_num] = comp; } } else if (a == 0) { for (comp_num = 0; comp_num < num_comp; comp_num++) { cv[comp_num] = bg; } } else { for (comp_num = 0; comp_num < num_comp; comp_num++) { comp = buf_ptr[x + planestride * comp_num]; cv[comp_num] = (comp << 8) + comp; } } color = dev_proc(target, encode_color)(target, cv); code = dev_proc(target, fill_rectangle)(target, x + x0, y + y0, 1, 1, color); } buf_ptr += rowstride; } return code; }