Merge pull request #5943 from lilyball/string_collect

Add new `string` subcommand `string collect`

.editorconfig

@@ -12,7 +12,7 @@ max_line_length = 100
 [{Makefile,*.in}]
 indent_style = tab
 
-[*.md]
+[*.{md,rst}]
 trim_trailing_whitespace = false
 
 [*.{sh,ac}]

CHANGELOG.md

@@ -15,6 +15,7 @@
 - The `--debug` option has been extended to allow specifying categories. Categories may be listed via `fish --print-debug-categories`.
 - `string replace` had an additional round of escaping in the replacement (not the match!), so escaping backslashes would require `string replace -ra '([ab])' '\\\\\\\$1' a`. A new feature flag `string-replace-fewer-backslashes` can be used to disable this, so that it becomes `string replace -ra '([ab])' '\\\\$1' a` (#5556).
 - Some parser errors did not set `$status` to non-zero. This has been corrected (b2a1da602f79878f4b0adc4881216c928a542608).
+- `string` has a new `collect` subcommand that disables newline-splitting on its input. This is meant to be used as the end of a command substitution pipeline to produce a single output argument potentially containing internal newlines, such as `set output (some-cmd | string collect)`. Any trailing newlines are trimmed, just like `"$(cmd)"` substitution in sh. It also supports a `--no-trim-newlines` flag to disable trailing newline trimming, which may be useful when doing something like `set contents (cat filename | string collect -N)` (#159).
 
 ### Syntax changes and new commands
 - Brace expansion now only takes place if the braces include a "," or a variable expansion, so things like `git reset HEAD@{0}` now work (#5869).

share/completions/string.fish

@@ -1,7 +1,7 @@
 # Completion for builtin string
 # This follows a strict command-then-options approach, so we can just test the number of tokens
 complete -f -c string
-complete -f -c string -n "test (count (commandline -opc)) -ge 2; and not contains -- (commandline -opc)[2] escape" -s q -l quiet -d "Do not print output"
+complete -f -c string -n "test (count (commandline -opc)) -ge 2; and not contains -- (commandline -opc)[2] escape collect" -s q -l quiet -d "Do not print output"
 complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "lower"
 complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "upper"
 complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "length"
@@ -13,6 +13,8 @@ complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "split0"
 complete -x -c string -n 'test (count (commandline -opc)) -ge 2; and string match -qr split0\?\$ -- (commandline -opc)[2]' -s m -l max -a "(seq 1 10)" -d "Specify maximum number of splits"
 complete -f -c string -n 'test (count (commandline -opc)) -ge 2; and string match -qr split0\?\$ -- (commandline -opc)[2]' -s r -l right -d "Split right-to-left"
 complete -f -c string -n 'test (count (commandline -opc)) -ge 2; and string match -qr split0\?\$ -- (commandline -opc)[2]' -s n -l no-empty -d "Empty results excluded"
+complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "collect"
+complete -f -c string -n 'test (count (commandline -opc)) -ge 2; and string match -qr collect\$ -- (commandline -opc)[2]' -s N -l no-trim-newlines -d "Don't trim trailing newlines"
 
 complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "join"
 complete -f -c string -n "test (count (commandline -opc)) -lt 2" -a "join0"

sphinx_doc_src/cmds/string.rst

@@ -6,6 +6,8 @@ string - manipulate strings
 Synopsis
 --------
 
+``string collect [(-N | --no-trim-newlines)] [STRING...]``
+
 ``string escape [(-n | --no-quoted)] [--style=xxx] [STRING...]``
 
 ``string join [(-q | --quiet)] SEP [STRING...]``
@@ -48,6 +50,17 @@ Most subcommands accept a ``-q`` or ``--quiet`` switch, which suppresses the usu
 
 The following subcommands are available.
 
+"collect" subcommand
+--------------------
+
+``string collect [(-N | --no-trim-newlines)] [STRING...]``
+
+``string collect`` collects its input into a single output argument, without splitting the output when used in a command substitution. This is useful when trying to collect multiline output from another command into a variable. Exit status: 0 if any output argument is non-empty, or 1 otherwise.
+
+If invoked with multiple arguments instead of input, ``string collect`` preserves each argument separately, where the number of output arguments is equal to the number of arguments given to ``string collect``.
+
+Any trailing newlines on the input are trimmed, just as with ``"$(cmd)"`` substitution in sh. ``--no-trim-newlines`` can be used to disable this behavior, which may be useful when running a command such as ``set contents (cat filename | string collect -N)``.
+
 "escape" and "unescape" subcommands
 -----------------------------------
 
@@ -329,6 +342,22 @@ Examples
     a1_20b2__c_E6_85_A1
 
 
+
+
+::
+
+    >_ echo \"(echo one\ntwo\nthree | string collect)\"
+    "one
+    two
+    three
+    "
+    
+    >_ echo \"(echo one\ntwo\nthree | string collect -N)\"
+    "one
+    two
+    three"
+
+
 Match Glob Examples
 -------------------
 

src/builtin_string.cpp

@@ -157,6 +157,7 @@ typedef struct {  //!OCLINT(too many fields)
     bool start_valid = false;
     bool style_valid = false;
     bool no_empty_valid = false;
+    bool no_trim_newlines_valid = false;
 
     bool all = false;
     bool entire = false;
@@ -171,6 +172,7 @@ typedef struct {  //!OCLINT(too many fields)
     bool regex = false;
     bool right = false;
     bool no_empty = false;
+    bool no_trim_newlines = false;
 
     long count = 0;
     long length = 0;
@@ -214,6 +216,9 @@ static int handle_flag_N(wchar_t **argv, parser_t &parser, io_streams_t &streams
     if (opts->no_newline_valid) {
         opts->no_newline = true;
         return STATUS_CMD_OK;
+    } else if (opts->no_trim_newlines_valid) {
+        opts->no_trim_newlines = true;
+        return STATUS_CMD_OK;
     }
     string_unknown_option(parser, streams, argv[0], argv[w.woptind - 1]);
     return STATUS_INVALID_ARGS;
@@ -408,12 +413,13 @@ static wcstring construct_short_opts(options_t *opts) {  //!OCLINT(high npath co
     if (opts->right_valid) short_opts.append(L"r");
     if (opts->start_valid) short_opts.append(L"s:");
     if (opts->no_empty_valid) short_opts.append(L"n");
+    if (opts->no_trim_newlines_valid) short_opts.append(L"N");
     return short_opts;
 }
 
 // Note that several long flags share the same short flag. That is okay. The caller is expected
 // to indicate that a max of one of the long flags sharing a short flag is valid.
-// Remember: adjust share/functions/string.fish when `string` options change
+// Remember: adjust share/completions/string.fish when `string` options change
 static const struct woption long_options[] = {
     {L"all", no_argument, NULL, 'a'},         {L"chars", required_argument, NULL, 'c'},
     {L"count", required_argument, NULL, 'n'}, {L"entire", no_argument, NULL, 'e'},
@@ -424,7 +430,8 @@ static const struct woption long_options[] = {
     {L"no-newline", no_argument, NULL, 'N'},  {L"no-quoted", no_argument, NULL, 'n'},
     {L"quiet", no_argument, NULL, 'q'},       {L"regex", no_argument, NULL, 'r'},
     {L"right", no_argument, NULL, 'r'},       {L"start", required_argument, NULL, 's'},
-    {L"style", required_argument, NULL, 1},   {NULL, 0, NULL, 0}};
+    {L"style", required_argument, NULL, 1},   {L"no-trim-newlines", no_argument, NULL, 'N'},
+    {NULL, 0, NULL, 0}};
 
 static const std::unordered_map<char, decltype(*handle_flag_N)> flag_to_function = {
     {'N', handle_flag_N}, {'a', handle_flag_a}, {'c', handle_flag_c}, {'e', handle_flag_e},
@@ -1125,6 +1132,29 @@ static int string_split0(parser_t &parser, io_streams_t &streams, int argc, wcha
     return string_split_maybe0(parser, streams, argc, argv, true /* is_split0 */);
 }
 
+static int string_collect(parser_t &parser, io_streams_t &streams, int argc, wchar_t **argv) {
+    options_t opts;
+    opts.no_trim_newlines_valid = true;
+    int optind;
+    int retval = parse_opts(&opts, &optind, 0, argc, argv, parser, streams);
+    if (retval != STATUS_CMD_OK) return retval;
+
+    auto &buff = streams.out.buffer();
+    arg_iterator_t aiter(argv, optind, streams, /* don't split */ false);
+    while (const wcstring *arg = aiter.nextstr()) {
+        auto begin = arg->cbegin(), end = arg->cend();
+        if (!opts.no_trim_newlines) {
+            while (end > begin && *(end-1) == L'\n') {
+                --end;
+            }
+        }
+
+        buff.append(begin, end, separation_type_t::explicitly);
+    }
+
+    return buff.size() > 0 ? STATUS_CMD_OK : STATUS_CMD_ERROR;
+}
+
 // Helper function to abstract the repeat logic from string_repeat
 // returns the to_repeat string, repeated count times.
 static wcstring wcsrepeat(const wcstring &to_repeat, size_t count) {
@@ -1305,7 +1335,8 @@ string_subcommands[] = {
     {L"length", &string_length}, {L"match", &string_match},       {L"replace", &string_replace},
     {L"split", &string_split},   {L"split0", &string_split0},     {L"sub", &string_sub},
     {L"trim", &string_trim},     {L"lower", &string_lower},       {L"upper", &string_upper},
-    {L"repeat", &string_repeat}, {L"unescape", &string_unescape}, {NULL, NULL}};
+    {L"repeat", &string_repeat}, {L"unescape", &string_unescape}, {L"collect", &string_collect},
+    {NULL, NULL}};
 
 /// The string builtin, for manipulating strings.
 int builtin_string(parser_t &parser, io_streams_t &streams, wchar_t **argv) {

tests/string.err

@@ -309,3 +309,9 @@ string repeat: Unknown option '-l'
 
 ####################
 # string split0 in functions
+
+####################
+# string collect
+
+####################
+# string collect in functions

tests/string.in

@@ -392,4 +392,31 @@ function dualsplit
 end
 count (dualsplit)
 
+logmsg string collect
+count (echo one\ntwo\nthree\nfour | string collect)
+count (echo one | string collect)
+echo [(echo one\ntwo\nthree | string collect)]
+echo [(echo one\ntwo\nthree | string collect -N)]
+printf '[%s]\n' (string collect one\n\n two\n)
+printf '[%s]\n' (string collect -N one\n\n two\n)
+printf '[%s]\n' (string collect --no-trim-newlines one\n\n two\n)
+# string collect returns 0 when it has any output, otherwise 1
+string collect >/dev/null; and echo unexpected success; or echo expected failure
+echo -n | string collect >/dev/null; and echo unexpected success; or echo expected failure
+echo | string collect -N >/dev/null; and echo expected success; or echo unexpected failure
+echo | string collect >/dev/null; and echo unexpected success; or echo expected failure
+string collect a >/dev/null; and echo expected success; or echo unexpected failure
+string collect -N '' >/dev/null; and echo unexpected success; or echo expected failure
+string collect \n\n >/dev/null; and echo unexpected success; or echo expected failure
+
+logmsg string collect in functions
+# This function outputs some newline-separated content, and some
+# explicitly un-separated content.
+function dualcollect
+  echo alpha
+  echo beta
+  echo gamma\ndelta\nomega | string collect
+end
+count (dualcollect)
+
 exit 0

tests/string.out

@@ -483,3 +483,38 @@ Split something
 ####################
 # string split0 in functions
 4
+
+####################
+# string collect
+1
+1
+[one
+two
+three]
+[one
+two
+three
+]
+[one]
+[two]
+[one
+
+]
+[two
+]
+[one
+
+]
+[two
+]
+expected failure
+expected failure
+expected success
+expected failure
+expected success
+expected failure
+expected failure
+
+####################
+# string collect in functions
+3