Implement printf, dprintf, string.format

Docs/Lua-String-Formatting.md

@@ -0,0 +1,12 @@
+# Lua String Formatting
+
+Lua contains a builtin routine, string.format, which is similar to "printf."
+We have reimplemented this from scratch, adding a bunch of new formatting
+directives.
+
+We have also added formatting routines that output to the
+console and the debug logs.  The new routines are:
+
+format(format_string, arg1, arg2, arg3...)
+dformat(format_string, arg1, arg2, arg3...)
+

luprex/cpp/core/luastack.hpp

@@ -700,6 +700,7 @@ public:
     // LUA_TTHREAD, LUA_TLIGHTUSERDATA, LUA_TUSERDATA.
     //
     int type(LuaSlot s) const { return lua_type(L_, s); }
+    const char *typestr(LuaSlot s) const { return lua_typename(L_, lua_type(L_, s)); }
 
     // Get the extended type of a LuaSlot.
     //

luprex/cpp/core/pprint.cpp

@@ -5,6 +5,7 @@
 #include "table.hpp"
 
 #include <cmath>
+#include <cinttypes>
 
 class PrintMachine {
 public:
@@ -185,24 +186,21 @@ public:
             }
         }
 
-        // If it's an array of atomic values, without a visible metatable,
-        // we're going to print it without newlines.  Scan the table to see if
-        // it's possible to print it array-style.
-        bool array_style = false;
-        if (!print_meta) {
-            array_style = true;
+        // Count the number of array-style keys.
+        // Also, check if there are any tables in the array keys.
+        int narray_keys = 0;
+        bool array_contains_table = false;
         for (int i = 1; i <= nkeys; i++) {
             LS.rawget(val, value, i);
-                if (LS.isnil(val) || (LS.type(val) == LUA_TTABLE)) {
-                    array_style = false;
-                    break;
-                }
-            }
+            if (LS.isnil(val)) break;
+            narray_keys = i;
+            if (LS.type(val) == LUA_TTABLE) array_contains_table = true;
         }
         
-        // Print it array-style.  This code is simple because
+        // Maybe print it array-style.  This code is simple because
         // we don't do indentation, and we don't handle any values
         // that aren't atomic.
+        bool array_style = (narray_keys == nkeys) && (!array_contains_table);
         if (array_style) {
             (*output_) << "{";
             for (int i = 1; i <= nkeys; i++) {
@@ -310,61 +308,231 @@ void pprint(LuaCoreStack &LS, LuaSlot val, const PrettyPrintOptions &opts, std::
     PrintMachine pm(LS, val, opts.indent, opts.level, opts.expand, os);
 }
 
-LuaDefine(string_pprint, "obj1, obj2, ...",
-    "|Pretty-print the specified objects into a string."
-    "|"
-    "|See also: string.pprintx, which has a lot more options."
-    "|This function uses the default options: pretty print indented,"
-    "|start at indentation level zero, and always expand the"
-    "|top-level table."
-    "|") {
-    LuaRet result;
-    LuaExtraArgs extra;
-    LuaDefStack LS(L, result, extra);
 
-    util::ostringstream oss;
-    for (int i = 0; i < extra.size(); i++) {
-        pprint(LS, extra[i], PrettyPrintOptions(), &oss);
-        oss << "\n";
+//////////////////////////////////////////////////////////////////////////////////
+//
+// Format 
+//
+// Printf-style formatting that consumes arguments from LuaExtraArgs.
+//
+//////////////////////////////////////////////////////////////////////////////////
+
+class FormatDirective
+{
+    // Given a string_view that starts with '%', count the number of characters
+    // in the format parameters: the '%', flags, width, and precision.  Does NOT
+    // include the conversion character.
+    //
+    // For example, given "%8.2d %2.7d", returns 4 (the length of "%8.2").
+    //
+    static int format_parameters_length(std::string_view fmt) {
+        assert ((fmt.size() >= 1) && (fmt[0] == '%'));
+        size_t i = 1;
+
+        // Flags
+        while (i < fmt.size() && (fmt[i] == '-' || fmt[i] == '+' || fmt[i] == ' ' || fmt[i] == '#' || fmt[i] == '0'))
+            i++;
+
+        // Width
+        while (i < fmt.size() && fmt[i] >= '0' && fmt[i] <= '9')
+            i++;
+
+        // Precision
+        if (i < fmt.size() && fmt[i] == '.') {
+            i++;
+            while (i < fmt.size() && fmt[i] >= '0' && fmt[i] <= '9')
+                i++;
+        }
+        return (int)i;
+    }
+
+public:
+    std::string_view precedingliteral;
+    std::string_view parameters;
+    std::string_view modifiers;
+    char directive;
+    char rebuilt[100];
+    const int PARAMETERS_TOO_LONG = 50;
+
+    // Return an error message declaring this format specifier to be invalid.
+    //
+    eng::string invalid()
+    {
+        return util::ss("Invalid format specifier: '", parameters, modifiers, directive, "'");
+    }
+
+    // Rebuild the format directive, using the specified suffix
+    // instead of the stored modifiers and directive.
+    //
+    const char *rebuild(std::string_view suffix)
+    {
+        std::string_view params = parameters;
+        if (int(params.size()) > PARAMETERS_TOO_LONG) params = "%";
+        memcpy(rebuilt, params.data(), params.size());
+        memcpy(rebuilt + params.size(), suffix.data(), suffix.size());
+        rebuilt[params.size() + suffix.size()] = 0;
+        return rebuilt;
+    }
+
+    // Read one directive from fmt, advancing fmt past everything consumed.
+    //
+    // On return:
+    //   directive != 0, parameters non-empty  — normal format directive
+    //   directive != 0, parameters empty       — not possible
+    //   directive == 0, parameters empty       — end of string (may have precedingliteral)
+    //   directive == 0, parameters non-empty   — truncated format (missing conversion char)
+    //
+    void read(std::string_view &fmt)
+    {
+        // Find the preceding literal (everything before the first '%').
+        size_t pct = fmt.find('%');
+        if (pct == std::string_view::npos) {
+            precedingliteral = fmt;
+            parameters = {};
+            modifiers = {};
+            directive = 0;
+            fmt = {};
+            return;
+        }
+
+        precedingliteral = fmt.substr(0, pct);
+        fmt.remove_prefix(pct);
+
+        // Measure format parameters (%, flags, width, precision).
+        int plen = format_parameters_length(fmt);
+        parameters = fmt.substr(0, plen);
+        fmt.remove_prefix(plen);
+
+        // Read 'l' modifiers.
+        int modcount = 0;
+        while ((modcount < int(fmt.size())) && (fmt[modcount] == 'l'))
+            modcount++;
+        modifiers = fmt.substr(0, modcount);
+        fmt.remove_prefix(modcount);
+
+        // Read conversion character.
+        if (fmt.empty()) { directive = 0; return; }
+        directive = fmt[0];
+        fmt.remove_prefix(1);
+    }
+};
+
+
+static void format_signed(LuaCoreStack &LS, LuaSlot arg, const char *format, std::ostream *os) {
+    auto ni = LS.tryinteger(arg);
+    if (ni) {
+        char buf[64];
+        snprintf(buf, sizeof(buf), format, *ni);
+        (*os) << buf;
     }
-    LS.set(result, oss.view());
-    return LS.result();
 }
 
-LuaDefine(string_pprintx, "options",
-    "|Pretty-print the specified object into a string, with options"
-    "|"
-    "|Options is a table with these fields:"
-    "|"
-    "|   value    - the object to pretty-print"
-    "|   indent   - if false, suppress newlines and indentation (default: true)"
-    "|   level    - base level of indentation (default: zero)"
-    "|   expand   - if true, force expansion of top-level table (default: false)"
-    "|"
-    "|About the expand flag: normally, when you print a class, it just "
-    "|prints '<class name>', and when you print a tangible, it just"
-    "|prints '<tangible id>'.  But sometimes, you want to see the details."
-    "|The expand flag forces it to expand the top-level table, even if the"
-    "|top-level table is a tangible or class."
-    "|") {
-    LuaArg loptions;
-    LuaRet result;
-    LuaVar value;
-    LuaDefStack LS(L, loptions, result, value);
-    PrettyPrintOptions options;
-    LuaKeywordParser kp(LS, loptions);
-    options.parse(kp);
-    if (!kp.optional(value, "value")) {
-        LS.set(value, LuaNil);
+static void format_unsigned(LuaCoreStack &LS, LuaSlot arg, const char *format, std::ostream *os) {
+    auto ni = LS.tryinteger(arg);
+    if (ni) {
+        char buf[64];
+        snprintf(buf, sizeof(buf), format, (uint64_t)(*ni));
+        (*os) << buf;
     }
-    kp.final_check_throw();
-    util::ostringstream oss;
-    pprint(LS, value, options, &oss);
-    LS.set(result, oss.view());
-    return LS.result();
 }
 
-LuaDefine(string_print, "obj",
+static void format_double(LuaCoreStack &LS, LuaSlot arg, const char *format, std::ostream *os) {
+    auto ni = LS.trynumber(arg);
+    if (ni) {
+        char buf[64];
+        snprintf(buf, sizeof(buf), format, *ni);
+        (*os) << buf;
+    }
+}
+
+
+eng::string format(LuaCoreStack &LS, std::string_view fmt, LuaExtraArgs args, std::ostream *os) {
+    FormatDirective fd;
+
+    // First pass: validate the format string and the argument types.
+    std::string_view fmtcopy = fmt;
+    int nargs = 0;
+    while (!fmtcopy.empty()) {
+        fd.read(fmtcopy);
+
+        if (int(fd.parameters.size()) > fd.PARAMETERS_TOO_LONG) return fd.invalid();
+
+        if (fd.directive == 0 && !fd.parameters.empty()) return fd.invalid();
+
+        if (fd.directive == '%') {
+            if ((fd.parameters.size() != 1)||(fd.modifiers.size() != 0))
+                return fd.invalid();
+        } else if (fd.directive != 0) {
+            if (nargs >= args.size())
+                return util::ss("expected more than ", args.size(), " arguments");
+            LuaSpecial arg = args[nargs++];
+            switch (fd.directive) {
+            case 'd': case 'i': case 'o': case 'u': case 'x': case 'X':
+            case 'c':
+                if (!LS.isinteger(arg))
+                    return util::ss("bad argument #", nargs, " (integer expected, got ", LS.typestr(arg), ")");
+                break;
+            case 'e': case 'E': case 'f': case 'g': case 'G':
+                if (!LS.isnumber(arg))
+                    return util::ss("bad argument #", nargs, " (number expected, got ", LS.typestr(arg), ")");
+                break;
+            case 's': case 'q': case 'p': case 'P':
+                break;
+            default:
+                return fd.invalid();
+            }
+        }
+    }
+
+    if (nargs != args.size())
+        return util::ss("expected ", nargs, " arguments, got ", args.size());
+
+    // Second pass: produce output.
+    int argidx = 0;
+    while (!fmt.empty()) {
+        fd.read(fmt);
+
+        if (!fd.precedingliteral.empty())
+            os->write(fd.precedingliteral.data(), fd.precedingliteral.size());
+
+        if (fd.directive == 0) break;
+        if (fd.directive == '%') { os->put('%'); continue; }
+
+        LuaSpecial arg = args[argidx++];
+
+        switch (fd.directive) {
+        case 'd': format_signed(LS, arg, fd.rebuild(PRId64), os); break;
+        case 'i': format_signed(LS, arg, fd.rebuild(PRId64), os); break;
+        case 'o': format_unsigned(LS, arg, fd.rebuild(PRIo64), os); break;
+        case 'u': format_unsigned(LS, arg, fd.rebuild(PRIu64), os); break;
+        case 'x': format_unsigned(LS, arg, fd.rebuild(PRIx64), os); break;
+        case 'X': format_unsigned(LS, arg, fd.rebuild(PRIX64), os); break;
+        case 'e': format_double(LS, arg, fd.rebuild("e"), os); break;
+        case 'E': format_double(LS, arg, fd.rebuild("E"), os); break;
+        case 'f': format_double(LS, arg, fd.rebuild("f"), os); break;
+        case 'g': format_double(LS, arg, fd.rebuild("g"), os); break;
+        case 'G': format_double(LS, arg, fd.rebuild("G"), os); break;
+        case 'c': format_signed(LS, arg, fd.rebuild("c"), os); break;
+        case 's': atomic_print(LS, arg, false, os); break;
+        case 'q': atomic_print(LS, arg, true, os); break;
+        case 'p': pprint(LS, arg, PrettyPrintOptions(fd.modifiers.empty(), false), os); break;
+        case 'P': pprint(LS, arg, PrettyPrintOptions(fd.modifiers.empty(), true),  os); break;
+        default: break;
+        }
+    }
+
+    return {};
+}
+
+//////////////////////////////////////////////////////////////////////////////////
+//
+// Lua Interfaces to the Various Printing Routines
+//
+// Note: there are more functions like this in world-accessor.
+//
+//////////////////////////////////////////////////////////////////////////////////
+
+LuaDefine(tostring, "obj",
     "|Concise print the specified object into a string"
     "|"
     "|This prints a concise representation of obj into a string.  Tables"
@@ -381,9 +549,9 @@ LuaDefine(string_print, "obj",
     return LS.result();
 }
 
-LuaDefineAlias(tostring, string_print);
 
-LuaDefine(string_isidentifier, "str", "return true if the string is a valid lua identifier") {
+LuaDefine(string_isidentifier, "str",
+    "return true if the string is a valid lua identifier") {
     LuaArg str;
     LuaRet result;
     LuaDefStack LS(L, str, result);
@@ -396,5 +564,3 @@ LuaDefine(string_isidentifier, "str", "return true if the string is a valid lua
     return LS.result();
 }
 
-
-

luprex/cpp/core/pprint.hpp

@@ -27,6 +27,7 @@ struct PrettyPrintOptions {
     int level;
     bool expand;
     PrettyPrintOptions() : indent(true), level(0), expand(true) {}
+    PrettyPrintOptions(bool indent, bool expand) : indent(indent), level(0), expand(expand) {}
     void parse(LuaKeywordParser &kp);
 };
 
@@ -42,4 +43,36 @@ void atomic_print(LuaCoreStack &LS, LuaSlot val, bool quote, std::ostream *os);
 //
 void pprint(LuaCoreStack &LS, LuaSlot val, const PrettyPrintOptions &opts, std::ostream *os);
 
+// Format a string using printf-style format specifiers, consuming
+// arguments from a LuaExtraArgs.  Returns an empty string on success,
+// or an error message on failure.
+//
+// Numeric types (argument must be a number).  All of the following
+// use the same formatting parameters as 'printf'.
+//   
+//   %d, %i   — signed decimal integer
+//   %o       — unsigned octal
+//   %u       — unsigned decimal
+//   %x, %X   — unsigned hexadecimal (lower/upper)
+//   %e, %E   — scientific notation (lower/upper)
+//   %f       — decimal floating point
+//   %g, %G   — shortest of %e/%f (lower/upper)
+//   %c       — character (integer converted to character)
+//
+// Non-numeric types:
+//   %s       — prints any lua value using atomic_print (unquoted)
+//   %q       — prints any lua value using atomic_print (quoted)
+//
+// Pretty-printing:
+//
+//   %p       - pretty print
+//   %lp      - pretty print, but all on one line
+//   %P       - pretty print, force table expansion
+//   %lP      - pretty print, force table expansion, all on one line
+//    
+// Special:
+//   %%       — literal percent sign (consumes no argument)
+//
+eng::string format(LuaCoreStack &LS, std::string_view fmt, LuaExtraArgs args, std::ostream *os);
+
 #endif // PPRINT_HPP

luprex/cpp/core/source.cpp

@@ -800,7 +800,6 @@ LuaDefineBuiltin(string_rep, "str, count", "repeat the string some number of tim
 LuaDefineBuiltin(string_reverse, "str", "reverse the bytes of the string");
 LuaDefineBuiltin(string_lower, "str", "convert string to lowercase");
 LuaDefineBuiltin(string_upper, "str", "convert string to uppercase");
-LuaDefineBuiltin(string_format, "formatstr, v1,v2,v3...", "generate formatted output string");
 LuaDefineBuiltin(string_gmatch, "str, pattern", "iterate over pattern-matched substrings");
 LuaDefineBuiltin(string_gsub, "str, pattern, replace", "global replace pattern in string");
 LuaDefineBuiltin(string_match, "str, pattern", "return start and end of pattern in string");

luprex/cpp/core/world-accessor.cpp

@@ -857,57 +857,6 @@ LuaDefine(math_randomstate, "seed",
 
 LuaSandboxBuiltin(math_randomseed, "", "");
 
-LuaDefine(pprint, "obj1, obj2, ...",
-    "|Pretty-print the specified objects."
-    "|"
-    "|See also: pprintx, which has a lot more options."
-    "|This function uses the default options: pretty print indented,"
-    "|start at indentation level zero, and always expand the"
-    "|top-level table."
-    "|") {
-    World *w = World::fetch_global_pointer(L);
-    std::ostream *ostream = w->lthread_print_stream();
-    LuaExtraArgs extra;
-    LuaDefStack LS(L, extra);
-    for (int i = 0; i < extra.size(); i++) {
-        pprint(LS, extra[i], PrettyPrintOptions(), ostream);
-        (*ostream) << std::endl;
-    }
-    return LS.result();
-}
-
-LuaDefine(pprintx, "options",
-    "|Pretty-print the specified object, with options"
-    "|"
-    "|Options is a table with these fields:"
-    "|"
-    "|   value    - the object to pretty-print"
-    "|   indent   - if false, suppress newlines and indentation (default: true)"
-    "|   level    - base level of indentation (default: zero)"
-    "|   expand   - if true, force expansion of top-level table (default: true)"
-    "|"
-    "|About the expand flag: normally, when you print a class, it just "
-    "|prints '<class name>', and when you print a tangible, it just"
-    "|prints '<tangible id>'.  But sometimes, you want to see the details."
-    "|The expand flag forces it to expand the top-level table, even if the"
-    "|top-level table is a tangible or class."
-    "|") {
-    World *w = World::fetch_global_pointer(L);
-    std::ostream *ostream = w->lthread_print_stream();
-    LuaArg loptions;
-    LuaVar value;
-    LuaDefStack LS(L, loptions, value);
-    PrettyPrintOptions options;
-    LuaKeywordParser kp(LS, loptions);
-    options.parse(kp);
-    if (!kp.optional(value, "value")) {
-        LS.set(value, LuaNil);
-    }
-    kp.final_check_throw();
-    pprint(LS, value, options, ostream);
-    return LS.result();
-}
-
 LuaDefine(print, "obj1, obj2, ...", 
     "|Print object or objects.") {
     World *w = World::fetch_global_pointer(L);
@@ -938,6 +887,88 @@ LuaDefine(dprint, "obj1, obj2, ...",
     return 0;
 }
 
+#define PRINTF_DOCS \
+    "|" \
+    "|Arguments: format, arg1, arg2, arg3..." \
+    "|" \
+    "|Numeric types (argument must be a number):" \
+    "|  %d, %i   -- signed decimal integer" \
+    "|  %o       -- unsigned octal" \
+    "|  %u       -- unsigned decimal" \
+    "|  %x, %X   -- unsigned hexadecimal (lower/upper)" \
+    "|  %e, %E   -- scientific notation (lower/upper)" \
+    "|  %f       -- decimal floating point" \
+    "|  %g, %G   -- shortest of %e/%f (lower/upper)" \
+    "|  %c       -- character (integer converted to character)" \
+    "|" \
+    "|Print any lua value:" \
+    "|  %s       -- Print unquoted" \
+    "|  %q       -- Print quoted" \
+    "|" \
+    "|Pretty-print any lua value, showing contents of tables:" \
+    "|  %p       -- pretty print" \
+    "|  %lp      -- pretty print, all on one line" \
+    "|  %P       -- pretty print, force table expansion" \
+    "|  %lP      -- pretty print, force table expansion, all on one line" \
+    "|" \
+    "|Other:" \
+    "|  %%       -- literal percent sign" \
+    "|"
+
+LuaDefine(printf, "fmt, ...",
+    "|Print a formatted string to the console."
+    "|"
+    PRINTF_DOCS) {
+    World *w = World::fetch_global_pointer(L);
+    std::ostream *ostream = w->lthread_print_stream();
+    LuaArg lfmt;
+    LuaExtraArgs extra;
+    LuaDefStack LS(L, lfmt, extra);
+    eng::string fmtstr = LS.ckstring(lfmt, "format string");
+    eng::string err = format(LS, fmtstr, extra, ostream);
+    if (!err.empty()) {
+        luaL_error(L, "%s", err.c_str());
+    }
+    (*ostream) << std::endl;
+    return 0;
+}
+
+LuaDefine(dprintf, "fmt, ...",
+    "|Print a formatted string to the debug log."
+    "|"
+    PRINTF_DOCS) {
+    LuaArg lfmt;
+    LuaExtraArgs extra;
+    LuaDefStack LS(L, lfmt, extra);
+    eng::string fmtstr = LS.ckstring(lfmt, "format string");
+    std::ostringstream oss;
+    eng::string err = format(LS, fmtstr, extra, &oss);
+    if (!err.empty()) {
+        luaL_error(L, "%s", err.c_str());
+    }
+    oss << std::endl;
+    util::dprintview(oss.str());
+    return 0;
+}
+
+LuaDefine(string_format, "fmt, ...",
+    "|Format a string using printf-style format specifiers."
+    "|"
+    PRINTF_DOCS) {
+    LuaArg lfmt;
+    LuaRet result;
+    LuaExtraArgs extra;
+    LuaDefStack LS(L, lfmt, result, extra);
+    eng::string fmtstr = LS.ckstring(lfmt, "format string");
+    util::ostringstream oss;
+    eng::string err = format(LS, fmtstr, extra, &oss);
+    if (!err.empty()) {
+        luaL_error(L, "%s", err.c_str());
+    }
+    LS.set(result, oss.view());
+    return LS.result();
+}
+
 LuaDefine(doc, "function",
     "|Print documentation for specified function.") {
     World *w = World::fetch_global_pointer(L);

luprex/cpp/core/world-core.cpp

@@ -1135,7 +1135,7 @@ void World::clear_lthread_state() {
     lthread_place_id_ = 0;
     lthread_thread_id_ = 0;
     lthread_use_ppool_ = false;
-    lthread_prints_.str();
+    lthread_prints_.str("");
     lthread_prints_.clear();
 }
 
@@ -1162,7 +1162,7 @@ void World::open_lthread_state(int64_t actor, int64_t place, int64_t thread, boo
     lthread_place_id_ = place;
     lthread_thread_id_ = thread;
     lthread_use_ppool_ = ppool;
-    lthread_prints_.str();
+    lthread_prints_.str("");
     lthread_prints_.clear();
 }