fix: escape CLI span JSON strings

tests/CMakeLists.txt

@@ -14,6 +14,11 @@ pf_add_test(test_graph_blocks)
 pf_add_test(test_ner)
 pf_add_test(test_tokenizer)
 
+# CLI serialization needs neither model weights nor ggml.
+add_executable(test_json test_json.cpp)
+target_compile_options(test_json PRIVATE -Wall -Wextra)
+add_test(NAME test_json COMMAND test_json)
+
 pf_add_test(test_parity)
 pf_add_test(test_window_stitch)
 set_tests_properties(test_parity test_window_stitch test_tokenizer

tests/test_json.cpp

@@ -0,0 +1,33 @@
+// SPDX-License-Identifier: MIT
+#include "../tools/json.h"
+
+#include <cstdio>
+#include <string>
+#include <string_view>
+
+int main() {
+    struct test_case { std::string input; std::string expected; };
+    const test_case cases[] = {
+        {"", "\"\""},
+        {"plain / text", "\"plain / text\""},
+        {"\"André\"", "\"\\\"André\\\"\""},
+        {"path\\file\tline\nnext\r\b\f", "\"path\\\\file\\tline\\nnext\\r\\b\\f\""},
+        {std::string("a\0b", 3), "\"a\\u0000b\""},
+        {std::string("\x01\x1f", 2), "\"\\u0001\\u001f\""},
+    };
+    for (const auto & c : cases) {
+        if (pf::cli::json_quote(c.input) != c.expected) {
+            std::fprintf(stderr, "JSON escaping mismatch\n");
+            return 1;
+        }
+    }
+    // Use an explicit span length: do not consume the remainder of the input.
+    if (pf::cli::json_quote(std::string_view("quoted\"tail", 6)) != "\"quoted\"") return 1;
+    // Every ASCII control byte must have an escaped representation.
+    for (int i = 0; i < 32; ++i) {
+        const char c = static_cast<char>(i);
+        const std::string quoted = pf::cli::json_quote(std::string_view(&c, 1));
+        if (quoted.size() < 4 || quoted[1] != '\\') return 1;
+    }
+    return 0;
+}

tools/json.h

@@ -0,0 +1,36 @@
+// SPDX-License-Identifier: MIT
+#pragma once
+
+#include <string>
+#include <string_view>
+
+namespace pf::cli {
+
+// Preserve UTF-8 bytes and escape JSON syntax and ASCII control characters.
+inline std::string json_quote(std::string_view text) {
+    static constexpr char hex[] = "0123456789abcdef";
+    std::string out = "\"";
+    for (unsigned char c : text) {
+        switch (c) {
+            case '"':  out += "\\\""; break;
+            case '\\': out += "\\\\"; break;
+            case '\b': out += "\\b"; break;
+            case '\f': out += "\\f"; break;
+            case '\n': out += "\\n"; break;
+            case '\r': out += "\\r"; break;
+            case '\t': out += "\\t"; break;
+            default:
+                if (c < 0x20) {
+                    out += "\\u00";
+                    out += hex[c >> 4];
+                    out += hex[c & 0x0f];
+                } else {
+                    out += static_cast<char>(c);
+                }
+        }
+    }
+    out += '"';
+    return out;
+}
+
+} // namespace pf::cli

tools/pf-cli.cpp

@@ -7,6 +7,7 @@
 #include "model.h"
 #include "ner.h"
 #include "tokenizer.h"
+#include "json.h"
 
 #include <cstdio>
 #include <cstring>
@@ -147,10 +148,12 @@ int main(int argc, char ** argv) {
         }
         std::printf("[");
         for (size_t i = 0; i < n_ents; i++) {
-            std::printf("%s\n {\"entity_group\": \"%s\", \"start\": %d, \"end\": %d, "
-                        "\"score\": %.4f, \"text\": \"%.*s\"}",
-                        i ? "," : "", ents[i].label, ents[i].start, ents[i].end,
-                        ents[i].score, ents[i].end - ents[i].start, text.data() + ents[i].start);
+            std::printf("%s\n {\"entity_group\": %s, \"start\": %d, \"end\": %d, "
+                        "\"score\": %.4f, \"text\": %s}",
+                        i ? "," : "", pf::cli::json_quote(ents[i].label).c_str(),
+                        ents[i].start, ents[i].end, ents[i].score,
+                        pf::cli::json_quote(std::string_view(text.data() + ents[i].start,
+                            ents[i].end - ents[i].start)).c_str());
         }
         std::printf("\n]\n");
         pf_entities_free(ents, n_ents);
@@ -250,10 +253,12 @@ int main(int argc, char ** argv) {
             while (bstart < bend && (unsigned char) text[bstart] <= ' ') bstart++;
             while (bend > bstart && (unsigned char) text[bend - 1] <= ' ') bend--;
             if (bstart >= bend) continue;
-            std::printf("%s\n {\"entity_group\": \"%s\", \"start\": %d, \"end\": %d, "
-                        "\"score\": %.4f, \"text\": \"%.*s\"}",
-                        first ? "" : ",", lt.category_name(sp.cat).c_str(), bstart, bend,
-                        sp.score, bend - bstart, text.data() + bstart);
+            std::printf("%s\n {\"entity_group\": %s, \"start\": %d, \"end\": %d, "
+                        "\"score\": %.4f, \"text\": %s}",
+                        first ? "" : ",", pf::cli::json_quote(lt.category_name(sp.cat)).c_str(),
+                        bstart, bend, sp.score,
+                        pf::cli::json_quote(std::string_view(text.data() + bstart,
+                            bend - bstart)).c_str());
             first = false;
         }
         std::printf("\n]\n");