ncplane_putegc_yx() copies directly as opposed to going through putc() #797

USAGE.md

@@ -1626,6 +1626,21 @@ void cell_release(struct ncplane* n, cell* c);
 #define NCSTYLE_PROTECT   0x00010000ul
 #define NCSTYLE_ITALIC    0x01000000ul
 
+// copy the UTF8-encoded EGC out of the cell, whether simple or complex. the
+// result is not tied to the ncplane, and persists across erases / destruction.
+static inline char*
+cell_strdup(const struct ncplane* n, const cell* c){
+  char* ret;
+  if(cell_simple_p(c)){
+    if( (ret = (char*)malloc(2)) ){ // cast is here for C++ clients
+      ret[0] = c->gcluster;
+      ret[1] = '\0';
+    }
+  }else{
+    ret = strdup(cell_extended_gcluster(n, c));
+  }
+  return ret;
+}
 
 // Set the specified style bits for the cell 'c', whether they're actively
 // supported or not.

include/notcurses/notcurses.h

@@ -729,7 +729,8 @@ cell_simple_p(const cell* c){
 // is invalidated by any further operation on the plane 'n', so...watch out!
 API const char* cell_extended_gcluster(const struct ncplane* n, const cell* c);
 
-// Extract the EGC from 'c' as a nul-terminated string.
+// copy the UTF8-encoded EGC out of the cell, whether simple or complex. the
+// result is not tied to the ncplane, and persists across erases / destruction.
 static inline char*
 cell_strdup(const struct ncplane* n, const cell* c){
   char* ret;

src/lib/internal.h

@@ -410,13 +410,6 @@ pool_egc_copy(const egcpool* e, const cell* c){
   return ret;
 }
 
-// copy the UTF8-encoded EGC out of the cell, whether simple or complex. the
-// result is not tied to the ncplane, and persists across erases / destruction.
-static inline char*
-cell_egc_copy(const ncplane* n, const cell* c){
-  return pool_egc_copy(&n->pool, c);
-}
-
 // For our first attempt, O(1) uniform conversion from 8-bit r/g/b down to
 // ~2.4-bit 6x6x6 cube + greyscale (assumed on entry; I know no way to
 // even semi-portably recover the palette) proceeds via: map each 8-bit to

src/lib/notcurses.c

@@ -1284,17 +1284,66 @@ int ncplane_putc_yx(ncplane* n, int y, int x, const cell* c){
 }
 
 int ncplane_putegc_yx(ncplane* n, int y, int x, const char* gclust, int* sbytes){
-  cell c = CELL_TRIVIAL_INITIALIZER;
-  int primed = cell_prime(n, &c, gclust, n->attrword, n->channels);
+  int cols;
+  int bytes = utf8_egc_len(gclust, &cols);
   if(sbytes){
-    *sbytes = primed;
+    *sbytes = bytes;
   }
-  if(primed < 0){
+  // if scrolling is enabled, check *before ncplane_cursor_move_yx()* whether
+  // we're past the end of the line, and move to the next line if so.
+  bool wide = cols > 1;
+  if(x == -1 && y == -1 && n->x + wide >= n->lenx){
+    if(!n->scrolling){
+      return -1;
+    }
+    scroll_down(n);
+  }
+  if(ncplane_cursor_move_yx(n, y, x)){
     return -1;
   }
-  int ret = ncplane_putc_yx(n, y, x, &c);
-  cell_release(n, &c);
-  return ret;
+  if(*gclust == '\n'){
+    if(n->scrolling){
+      scroll_down(n);
+      return 0;
+    }
+  }
+  // A wide character obliterates anything to its immediate right (and marks
+  // that cell as wide). Any character placed atop one half of a wide character
+  // obliterates the other half. Note that a wide char can thus obliterate two
+  // wide chars, totalling four columns.
+  cell* targ = ncplane_cell_ref_yx(n, n->y, n->x);
+  if(n->x > 0){
+    if(cell_double_wide_p(targ)){ // replaced cell is half of a wide char
+      if(targ->gcluster == 0){ // we're the right half
+        cell_obliterate(n, &n->fb[nfbcellidx(n, n->y, n->x - 1)]);
+      }else{
+        cell_obliterate(n, &n->fb[nfbcellidx(n, n->y, n->x + 1)]);
+      }
+    }
+  }
+  uint64_t channels = n->channels & ~CELL_WIDEASIAN_MASK;
+  if(wide){
+    channels |= CELL_WIDEASIAN_MASK;
+  }
+  if(cell_prime(n, targ, gclust, n->attrword, channels) < 0){
+    return -1;
+  }
+  if(wide){ // must set our right wide, and check for further damage
+    if(n->x < n->lenx - 1){ // check to our right
+      cell* candidate = &n->fb[nfbcellidx(n, n->y, n->x + 1)];
+      if(n->x < n->lenx - 2){
+        if(cell_wide_left_p(candidate)){
+          cell_obliterate(n, &n->fb[nfbcellidx(n, n->y, n->x + 2)]);
+        }
+      }
+      cell_obliterate(n, candidate);
+      cell_set_wide(candidate);
+      candidate->channels = channels;
+      candidate->attrword = n->attrword;
+    }
+  }
+  n->x += cols;
+  return cols;
 }
 
 int ncplane_putsimple_stainable(ncplane* n, char c){
@@ -1867,7 +1916,7 @@ void ncplane_erase(ncplane* n){
   // we must preserve the background, but a pure cell_duplicate() would be
   // wiped out by the egcpool_dump(). do a duplication (to get the attrword
   // and channels), and then reload.
-  char* egc = cell_egc_copy(n, &n->basecell);
+  char* egc = cell_strdup(n, &n->basecell);
   memset(n->fb, 0, sizeof(*n->fb) * n->lenx * n->leny);
   egcpool_dump(&n->pool);
   egcpool_init(&n->pool);

tests/wide.cpp

@@ -284,7 +284,7 @@ TEST_CASE("Wide") {
 
     // should be wide char 1
     REQUIRE(3 == ncplane_at_yx_cell(n_, 0, 0, &c));
-    egc = cell_egc_copy(n_, &c);
+    egc = cell_strdup(n_, &c);
     REQUIRE(egc);
     CHECK(!strcmp("\xe5\x85\xa8", egc));
     CHECK(cell_double_wide_p(&c));
@@ -297,7 +297,7 @@ TEST_CASE("Wide") {
     cell_init(&c);
     // should be wide char 1 right side
     REQUIRE(0 == ncplane_at_yx_cell(n_, 0, 1, &c));
-    egc = cell_egc_copy(n_, &c);
+    egc = cell_strdup(n_, &c);
     REQUIRE(egc);
     CHECK(!strcmp("", egc));
     CHECK(cell_double_wide_p(&c));
@@ -311,7 +311,7 @@ TEST_CASE("Wide") {
 
     // should be wide char 2
     REQUIRE(3 == ncplane_at_yx_cell(n_, 0, 2, &c));
-    egc = cell_egc_copy(n_, &c);
+    egc = cell_strdup(n_, &c);
     REQUIRE(egc);
     CHECK(!strcmp("\xe5\xbd\xa2", egc));
     CHECK(cell_double_wide_p(&c));
@@ -324,7 +324,7 @@ TEST_CASE("Wide") {
     cell_init(&c);
     // should be wide char 2 right side
     CHECK(0 == ncplane_at_yx_cell(n_, 0, 3, &c));
-    egc = cell_egc_copy(n_, &c);
+    egc = cell_strdup(n_, &c);
     REQUIRE(egc);
     CHECK(!strcmp("", egc));
     CHECK(cell_double_wide_p(&c));