complete/fish: improve shell completions for fish

- Stop using `-n __fish_use_subcommand`. This had the effect of
ignoring options if a positional argument has already been given, but
that's not how ripgrep works.

- Only suggest negation options if the option they're negating is
passed (e.g., only complete `--no-pcre2` if `--pcre2` is present). The
zsh completions already do this.

- Take into account whether an option takes an argument. If an option
is not a switch then it won't suggest further options until the
argument is given, e.g. `-C<tab>` won't suggest options but `-i<tab>`
will.

- Suggest correct arguments for options. We already completed a fixed
set of choices where available, but now we go further:

  - Filenames are only suggested for options that take filenames.

  - `--pre` and `--hostname-bin` suggest binaries from `$PATH`.

  - `-t`/`--type`/&c use `--type-list` for suggestions, like in zsh,
  with a preview of the glob patterns.

  - `--encoding` uses a hardcoded list extracted from the zsh
  completions. This has been refactored into a separate file, and the
  range globs (`{1..5}`) replaced by comma globs (`{1,2,3,4,5}`) since
  those work in both shells. I verified that this produces the same
  list as before in zsh, and the same list in fish (albeit in a
  different order).

PR #2684

8 files changed, 131 insertions, 52 deletions

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 366f904e..e6d56e7f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,13 @@
+14.1.0 (TBD)
+============
+This is a minor release with a few small new features and bug fixes.
+
+Feature enhancements:
+
+* [FEATURE #2684](https://github.com/BurntSushi/ripgrep/issues/2684):
+  Improve completions for the `fish` shell.
+
+
 14.0.3 (2023-11-28)
 ===================
 This is a patch release with a bug fix for the `--sortr` flag.
diff --git a/crates/core/flags/complete/encodings.sh b/crates/core/flags/complete/encodings.sh
new file mode 100644
index 00000000..14fc4a85
--- /dev/null
+++ b/crates/core/flags/complete/encodings.sh
@@ -0,0 +1,29 @@
+# This is impossible to read, but these encodings rarely if ever change, so
+# it probably does not matter. They are derived from the list given here:
+# https://encoding.spec.whatwg.org/#concept-encoding-get
+#
+# The globbing here works in both fish and zsh (though they expand it in
+# different orders). It may work in other shells too.
+
+{{,us-}ascii,arabic,chinese,cyrillic,greek{,8},hebrew,korean}
+logical visual mac {,cs}macintosh x-mac-{cyrillic,roman,ukrainian}
+866 ibm{819,866} csibm866
+big5{,-hkscs} {cn-,cs}big5 x-x-big5
+cp{819,866,125{0,1,2,3,4,5,6,7,8}} x-cp125{0,1,2,3,4,5,6,7,8}
+csiso2022{jp,kr} csiso8859{6,8}{e,i}
+csisolatin{1,2,3,4,5,6,9} csisolatin{arabic,cyrillic,greek,hebrew}
+ecma-{114,118} asmo-708 elot_928 sun_eu_greek
+euc-{jp,kr} x-euc-jp cseuckr cseucpkdfmtjapanese
+{,x-}gbk csiso58gb231280 gb18030 {,cs}gb2312 gb_2312{,-80} hz-gb-2312
+iso-2022-{cn,cn-ext,jp,kr}
+iso8859{,-}{1,2,3,4,5,6,7,8,9,10,11,13,14,15}
+iso-8859-{1,2,3,4,5,6,7,8,9,10,11,{6,8}-{e,i},13,14,15,16} iso_8859-{1,2,3,4,5,6,7,8,9,15}
+iso_8859-{1,2,6,7}:1987 iso_8859-{3,4,5,8}:1988 iso_8859-9:1989
+iso-ir-{58,100,101,109,110,126,127,138,144,148,149,157}
+koi{,8,8-r,8-ru,8-u,8_r} cskoi8r
+ks_c_5601-{1987,1989} ksc{,_}5691 csksc56011987
+latin{1,2,3,4,5,6} l{1,2,3,4,5,6,9}
+shift{-,_}jis csshiftjis {,x-}sjis ms_kanji ms932
+utf{,-}8 utf-16{,be,le} unicode-1-1-utf-8
+windows-{31j,874,949,125{0,1,2,3,4,5,6,7,8}} dos-874 tis-620 ansi_x3.4-1968
+x-user-defined auto none
diff --git a/crates/core/flags/complete/fish.rs b/crates/core/flags/complete/fish.rs
index e55d72e3..f8f7133b 100644
--- a/crates/core/flags/complete/fish.rs
+++ b/crates/core/flags/complete/fish.rs
@@ -2,17 +2,13 @@
 Provides completions for ripgrep's CLI for the fish shell.
 */
 
-use crate::flags::defs::FLAGS;
+use crate::flags::{defs::FLAGS, CompletionType};
 
-const TEMPLATE: &'static str =
-    "complete -c rg -n '__fish_use_subcommand' !SHORT! !LONG! !DOC!\n";
-const TEMPLATE_CHOICES: &'static str =
-    "complete -c rg -n '__fish_use_subcommand' !SHORT! !LONG! !DOC! -r -f -a '!CHOICES!'\n";
+const TEMPLATE: &'static str = "complete -c rg !SHORT! -l !LONG! -d '!DOC!'";
+const TEMPLATE_NEGATED: &'static str =
+    "complete -c rg -l !NEGATED! -n '__fish_contains_opt !SHORT! !LONG!' -d '!DOC!'\n";
 
 /// Generate completions for Fish.
-///
-/// Note that these completions are based on what was produced for ripgrep <=13
-/// using Clap 2.x. Improvements on this are welcome.
 pub(crate) fn generate() -> String {
     let mut out = String::new();
     for flag in FLAGS.iter() {
@@ -20,25 +16,49 @@ pub(crate) fn generate() -> String {
             None => "".to_string(),
             Some(byte) => format!("-s {}", char::from(byte)),
         };
-        let long = format!("-l '{}'", flag.name_long().replace("'", "\\'"));
-        let doc = format!("-d '{}'", flag.doc_short().replace("'", "\\'"));
-        let template = if flag.doc_choices().is_empty() {
-            TEMPLATE.to_string()
-        } else {
-            TEMPLATE_CHOICES
-                .replace("!CHOICES!", &flag.doc_choices().join(" "))
-        };
-        out.push_str(
-            &template
-                .replace("!SHORT!", &short)
-                .replace("!LONG!", &long)
-                .replace("!DOC!", &doc),
-        );
+        let long = flag.name_long();
+        let doc = flag.doc_short().replace("'", "\\'");
+        let mut completion = TEMPLATE
+            .replace("!SHORT!", &short)
+            .replace("!LONG!", &long)
+            .replace("!DOC!", &doc);
+
+        match flag.completion_type() {
+            CompletionType::Filename => {
+                completion.push_str(" -r -F");
+            }
+            CompletionType::Executable => {
+                completion.push_str(" -r -f -a '(__fish_complete_command)'");
+            }
+            CompletionType::Filetype => {
+                completion.push_str(
+                    " -r -f -a '(rg --type-list | string replace : \\t)'",
+                );
+            }
+            CompletionType::Encoding => {
+                completion.push_str(" -r -f -a '");
+                completion.push_str(super::ENCODINGS);
+                completion.push_str("'");
+            }
+            CompletionType::Other if !flag.doc_choices().is_empty() => {
+                completion.push_str(" -r -f -a '");
+                completion.push_str(&flag.doc_choices().join(" "));
+                completion.push_str("'");
+            }
+            CompletionType::Other if !flag.is_switch() => {
+                completion.push_str(" -r -f");
+            }
+            CompletionType::Other => (),
+        }
+
+        completion.push('\n');
+        out.push_str(&completion);
+
         if let Some(negated) = flag.name_negated() {
-            let long = format!("-l '{}'", negated.replace("'", "\\'"));
             out.push_str(
-                &TEMPLATE
-                    .replace("!SHORT!", "")
+                &TEMPLATE_NEGATED
+                    .replace("!NEGATED!", &negated)
+                    .replace("!SHORT!", &short)
                     .replace("!LONG!", &long)
                     .replace("!DOC!", &doc),
             );
diff --git a/crates/core/flags/complete/mod.rs b/crates/core/flags/complete/mod.rs
index 7d2fb606..b531bf12 100644
--- a/crates/core/flags/complete/mod.rs
+++ b/crates/core/flags/complete/mod.rs
@@ -2,6 +2,8 @@
 Modules for generating completions for various shells.
 */
 
+static ENCODINGS: &'static str = include_str!("encodings.sh");
+
 pub(super) mod bash;
 pub(super) mod fish;
 pub(super) mod powershell;
diff --git a/crates/core/flags/complete/rg.zsh b/crates/core/flags/complete/rg.zsh
index 7ca89008..e9731a16 100644
--- a/crates/core/flags/complete/rg.zsh
+++ b/crates/core/flags/complete/rg.zsh
@@ -413,32 +413,8 @@ _rg_encodings() {
   local -a expl
   local -aU _encodings
 
-  # This is impossible to read, but these encodings rarely if ever change, so it
-  # probably doesn't matter. They are derived from the list given here:
-  # https://encoding.spec.whatwg.org/#concept-encoding-get
   _encodings=(
-    {{,us-}ascii,arabic,chinese,cyrillic,greek{,8},hebrew,korean}
-    logical visual mac {,cs}macintosh x-mac-{cyrillic,roman,ukrainian}
-    866 ibm{819,866} csibm866
-    big5{,-hkscs} {cn-,cs}big5 x-x-big5
-    cp{819,866,125{0..8}} x-cp125{0..8}
-    csiso2022{jp,kr} csiso8859{6,8}{e,i}
-    csisolatin{{1..6},9} csisolatin{arabic,cyrillic,greek,hebrew}
-    ecma-{114,118} asmo-708 elot_928 sun_eu_greek
-    euc-{jp,kr} x-euc-jp cseuckr cseucpkdfmtjapanese
-    {,x-}gbk csiso58gb231280 gb18030 {,cs}gb2312 gb_2312{,-80} hz-gb-2312
-    iso-2022-{cn,cn-ext,jp,kr}
-    iso8859{,-}{{1..11},13,14,15}
-    iso-8859-{{1..11},{6,8}-{e,i},13,14,15,16} iso_8859-{{1..9},15}
-    iso_8859-{1,2,6,7}:1987 iso_8859-{3,4,5,8}:1988 iso_8859-9:1989
-    iso-ir-{58,100,101,109,110,126,127,138,144,148,149,157}
-    koi{,8,8-r,8-ru,8-u,8_r} cskoi8r
-    ks_c_5601-{1987,1989} ksc{,_}5691 csksc56011987
-    latin{1..6} l{{1..6},9}
-    shift{-,_}jis csshiftjis {,x-}sjis ms_kanji ms932
-    utf{,-}8 utf-16{,be,le} unicode-1-1-utf-8
-    windows-{31j,874,949,125{0..8}} dos-874 tis-620 ansi_x3.4-1968
-    x-user-defined auto none
+!ENCODINGS!
   )
 
   _wanted encodings expl encoding compadd -a "$@" - _encodings
diff --git a/crates/core/flags/complete/zsh.rs b/crates/core/flags/complete/zsh.rs
index 59876f18..7475c0c1 100644
--- a/crates/core/flags/complete/zsh.rs
+++ b/crates/core/flags/complete/zsh.rs
@@ -19,5 +19,5 @@ long as it meets criteria 3 and 4 above.
 
 /// Generate completions for zsh.
 pub(crate) fn generate() -> String {
-    include_str!("rg.zsh").to_string()
+    include_str!("rg.zsh").replace("!ENCODINGS!", super::ENCODINGS.trim_end())
 }
diff --git a/crates/core/flags/defs.rs b/crates/core/flags/defs.rs
index a0ace969..0abf0c92 100644
--- a/crates/core/flags/defs.rs
+++ b/crates/core/flags/defs.rs
@@ -34,6 +34,8 @@ use crate::flags::{
 #[cfg(test)]
 use crate::flags::parse::parse_low_raw;
 
+use super::CompletionType;
+
 /// A list of all flags in ripgrep via implementations of `Flag`.
 ///
 /// The order of these flags matter. It determines the order of the flags in
@@ -1582,6 +1584,9 @@ The encoding detection that ripgrep uses can be reverted to its automatic mode
 via the \flag-negate{encoding} flag.
 "
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Encoding
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         let value = match v {
@@ -1977,6 +1982,9 @@ When \flag{file} or \flag{regexp} is used, then ripgrep treats all positional
 arguments as files or directories to search.
 "
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Filename
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         let path = PathBuf::from(v.unwrap_value());
@@ -2808,6 +2816,9 @@ to calling \fBgethostname\fP. On Windows, this corresponds to calling
 ripgrep uses your system's hostname for producing hyperlinks.
 "#
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Executable
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         let path = PathBuf::from(v.unwrap_value());
@@ -3141,6 +3152,9 @@ If you are looking for a way to include or exclude files and directories
 directly on the command line, then use \flag{glob} instead.
 "
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Filename
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         let path = PathBuf::from(v.unwrap_value());
@@ -5410,6 +5424,9 @@ format, then \fBpzstd\fP is used to decompress the contents to stdout.
 This overrides the \flag{search-zip} flag.
 "#
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Executable
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         let path = match v {
@@ -6781,6 +6798,9 @@ any rules found in ignore files.
 To see the list of available file types, use the \flag{type-list} flag.
 "#
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Filetype
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         args.type_changes.push(TypeChange::Select {
@@ -7000,6 +7020,9 @@ will only search files that are unrecognized by its type definitions.
 To see the list of available file types, use the \flag{type-list} flag.
 "#
     }
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Filetype
+    }
 
     fn update(&self, v: FlagValue, args: &mut LowArgs) -> anyhow::Result<()> {
         args.type_changes.push(TypeChange::Negate {
diff --git a/crates/core/flags/mod.rs b/crates/core/flags/mod.rs
index e2b0ccb8..54aec572 100644
--- a/crates/core/flags/mod.rs
+++ b/crates/core/flags/mod.rs
@@ -70,7 +70,7 @@ mod parse;
 /// value. Flags that accept multiple values are an unsupported abberation.
 trait Flag: Debug + Send + Sync + UnwindSafe + RefUnwindSafe + 'static {
     /// Returns true if this flag is a switch. When a flag is a switch, the
-    /// CLI parser will look for a value after the flag is seen.
+    /// CLI parser will not look for a value after the flag is seen.
     fn is_switch(&self) -> bool;
 
     /// A short single byte name for this flag. This returns `None` by default,
@@ -150,6 +150,10 @@ trait Flag: Debug + Send + Sync + UnwindSafe + RefUnwindSafe + 'static {
         &[]
     }
 
+    fn completion_type(&self) -> CompletionType {
+        CompletionType::Other
+    }
+
     /// Given the parsed value (which might just be a switch), this should
     /// update the state in `args` based on the value given for this flag.
     ///
@@ -228,6 +232,21 @@ impl Category {
     }
 }
 
+/// The kind of argument a flag accepts, to be used for shell completions.
+#[derive(Clone, Copy, Debug)]
+enum CompletionType {
+    /// No special category. is_switch() and doc_choices() may apply.
+    Other,
+    /// A path to a file.
+    Filename,
+    /// A command in $PATH.
+    Executable,
+    /// The name of a file type, as used by e.g. --type.
+    Filetype,
+    /// The name of an encoding_rs encoding, as used by --encoding.
+    Encoding,
+}
+
 /// Represents a value parsed from the command line.
 ///
 /// This doesn't include the corresponding flag, but values come in one of