mandb: support the formats of the man pages of "awk" and "sed"

docs/ChangeLog.md

@@ -131,6 +131,7 @@
   - mandb: fix an infinite loop by a leak variable (reported by rlanore, riblo) `#D1550` 0efcb65
   - mandb: work around old groff in macOS (reported by killermoehre) `#D1551` d4f816b
   - mandb: use `manpath` and `man -w`, and read `/etc/man_db.conf` and `~/.manpath` `#D1637` 2365e09
+  - mandb: support the formats of the man pages of `awk` and `sed` (reported by bbyfacekiller) `#D1687` 0000000
 - edit: work around the wrong job information of Bash in trap handlers (reported by 3ximus) `#D1435` `#D1436` bc4735e
 - edit (command-help): work around the Bash bug that tempenv vanishes with `builtin eval` `#D1438` 8379d4a
 - global: suppress missing locale errors (reported by 3ximus) `#D1440` 4d3c595

lib/core-complete.sh

@@ -3379,48 +3379,83 @@ function ble/complete/mandb/.generate-cache {
       g_desc = "";
     }
 
-    mode == "key1" {
-      register_key($0);
-      mode = "desc";
+    # Comment: [.ig \n comments \n ..]
+    /^\.ig/ { mode = "ignore"; next; }
+    mode == "ignore" {
+      if (/^\.\.[[:space:]]*/) mode = "none";
       next;
     }
 
-    mode == "keyc" {
-      g_current_key = g_current_key "\n" $0;
-      if (/^\.Xc/) {
-        register_key(g_current_key);
-        mode = "desc";
-      }
-      next;
-    }
+    /^\.(S[Ss]|S[Hh]|P[Pp])([^_[:alnum:]]|$)/ { flush_topic(); next; }
 
-    type == "man" && /^\.TP([^_[:alnum:]]|$)/ {
+    #--------------------------------------------------------------------------
+    # Format #3: [.HP \n keys \n .IP \n desc]
+    # GNU sed seems to use this format.
+    /^\.HP[[:space:]]*$/ {
       if (g_keys_count && g_desc != "") flush_topic();
-      mode = "key1";
+      mode = "fmt3_key";
+    }
+    mode == "fmt3_key" {
+      if (/^\.TP[[:space:]]*/) { flush_topic(); mode = "none"; next; }
+      if (/^\.PD([^_[:alnum:]]|$)/) next;
+
+      if (/^\.IP/) { mode = "fmt3_desc"; next; }
+      register_key($0);
       next;
     }
+    mode == "fmt3_desc" {
+      if (/^\.TP[[:space:]]*/) { flush_topic(); mode = "none"; next; }
+      if (/^\.PD([^_[:alnum:]]|$)/) next;
 
-#%  # mdoc でも man でもあった
+      if (g_desc != "") g_desc = g_desc "\n";
+      g_desc = g_desc $0;
+      next;
+    }
+    #--------------------------------------------------------------------------
+    # Format #2: [.It Fl key \n desc] or [.It Fl Xo \n key \n .Xc desc]
+    # This form was found in both "mdoc" and "man"
     /^\.It Fl([^_[:alnum:]]|$)/ {
       if (g_keys_count && g_desc != "") flush_topic();
       sub(/^\.It Fl/, ".Fl");
       if ($0 ~ / Xo$/) {
         g_current_key = $0;
-        mode = "keyc"
+        mode = "fmt2_keyc"
       } else {
         register_key($0);
         mode = "desc";
       }
       next;
     }
-
-    /^\.(S[Ss]|S[Hh]|P[Pp])([^_[:alnum:]]|$)/ { flush_topic(); next; }
-    /^\.PD([^_[:alnum:]]|$)/ { next; }
-
+    mode == "fmt2_keyc" {
+      if (/^\.PD[[:space:]]*([0-9]+[[:space:]]*)?$/) next;
+      g_current_key = g_current_key "\n" $0;
+      if (/^\.Xc/) {
+        register_key(g_current_key);
+        mode = "desc";
+      }
+      next;
+    }
+    #--------------------------------------------------------------------------
+    # Format #1: [.TP \n key \n desc]
+    # This is the typical format in "man".
+    type == "man" && /^\.TP([^_[:alnum:]]|$)/ {
+      if (g_keys_count && g_desc != "") flush_topic();
+      mode = "key1";
+      next;
+    }
+    mode == "key1" {
+      if (/^\.PD[[:space:]]*([0-9]+[[:space:]]*)?$/) next;
+      register_key($0);
+      mode = "desc";
+      next;
+    }
     mode == "desc" {
+      if (/^\.PD([^_[:alnum:]]|$)/) next;
+
       if (g_desc != "") g_desc = g_desc "\n";
       g_desc = g_desc $0;
     }
+    #--------------------------------------------------------------------------
 
     END { flush_topic(); }
   ' | ble/complete/mandb/convert-mandoc 2>/dev/null | ble/bin/awk '

note.txt

@@ -1599,12 +1599,15 @@ bash_tips
   * return without arguments in trap handlers. これは結局強い理由がないと変更さ
     れない雰囲気になっている。
 
-
 @todo
 *******************************************************************************
     ToDo
 -------------------------------------------------------------------------------
 
+2021-12-06
+
+  * mandb: 空文字列の時にも mandb に基づくオプション生成を実行するべき
+
 2021-11-23
 
   * kitty shell-integration (requested by kovidgoyal)
@@ -5541,6 +5544,16 @@ bash_tips
 
 2021-12-06
 
+  * mandb: awk, sed の man ページからの抽出に失敗している (reported by bbyfacekiller) [#D1687]
+    https://github.com/akinomyoga/ble.sh/issues/152
+
+    当初は bash-completion との integration の問題かとも思ったが、それ以前に
+    mandb での抽出に於いて awk, sed でそれぞれ異なる形で失敗していた。両方につ
+    いて対応した。
+
+    更に、awk の man page の中にある ".ig ... .." はコメントセクションの様だ。
+    古いオプションや文章が含まれているらしい。
+
   * 2021-09-06 menu: fish の様に twocolumn 形式で表示する desc を作っても良いのではないか (motivated by Shahabaz-Bagwan) [#D1686]
     https://www.youtube.com/watch?v=YS1vxEhd2Pc
     https://github.com/akinomyoga/ble.sh/issues/132