skip UTF-8 BOM if present

include/simdjson/dom/document_stream.h

@@ -224,8 +224,7 @@ class document_stream {
    * Parse the next document found in the buffer previously given to document_stream.
    *
    * The content should be a valid JSON document encoded as UTF-8. If there is a
-   * UTF-8 BOM, the caller is responsible for omitting it, UTF-8 BOM are
-   * discouraged.
+   * UTF-8 BOM, the parser skips it.
    *
    * You do NOT need to pre-allocate a parser.  This function takes care of
    * pre-allocating a capacity defined by the batch_size defined when creating the

include/simdjson/dom/parser-inl.h

@@ -12,6 +12,7 @@
 #include "simdjson/dom/element-inl.h"
 
 #include <climits>
+#include <cstring> /* memcmp */
 
 namespace simdjson {
 namespace dom {
@@ -120,8 +121,14 @@ inline simdjson_result<element> parser::parse_into_document(document& provided_d
       _loaded_bytes_capacity = len;
     }
     std::memcpy(static_cast<void *>(loaded_bytes.get()), buf, len);
+    buf = reinterpret_cast<const uint8_t*>(loaded_bytes.get());
   }
-  _error = implementation->parse(realloc_if_needed ? reinterpret_cast<const uint8_t*>(loaded_bytes.get()): buf, len, provided_doc);
+
+  if((len >= 3) && (std::memcmp(buf, "\xEF\xBB\xBF", 3) == 0)) {
+    buf += 3;
+    len -= 3;
+  }
+  _error = implementation->parse(buf, len, provided_doc);
 
   if (_error) { return _error; }
 
@@ -158,6 +165,10 @@ simdjson_inline simdjson_result<element> parser::parse(const padded_string_view
 
 inline simdjson_result<document_stream> parser::parse_many(const uint8_t *buf, size_t len, size_t batch_size) noexcept {
   if(batch_size < MINIMAL_BATCH_SIZE) { batch_size = MINIMAL_BATCH_SIZE; }
+  if((len >= 3) && (std::memcmp(buf, "\xEF\xBB\xBF", 3) == 0)) {
+    buf += 3;
+    len -= 3;
+  }
   return document_stream(*this, buf, len, batch_size);
 }
 inline simdjson_result<document_stream> parser::parse_many(const char *buf, size_t len, size_t batch_size) noexcept {

include/simdjson/dom/parser.h

@@ -254,6 +254,8 @@ class parser {
    * And, possibly, no document many have been parsed when the `parser.load_many(path)` function
    * returned.
    *
+   * If there is a UTF-8 BOM, the parser skips it.
+   *
    * ### Format
    *
    * The file must contain a series of one or more JSON documents, concatenated into a single
@@ -346,6 +348,8 @@ class parser {
    *     cout << std::string(doc["title"]) << endl;
    *   }
    *
+   * If there is a UTF-8 BOM, the parser skips it.
+   *
    * ### Format
    *
    * The buffer must contain a series of one or more JSON documents, concatenated into a single

include/simdjson/generic/ondemand/document_stream.h

@@ -241,8 +241,7 @@ class document_stream {
    * Parse the next document found in the buffer previously given to document_stream.
    *
    * The content should be a valid JSON document encoded as UTF-8. If there is a
-   * UTF-8 BOM, the caller is responsible for omitting it, UTF-8 BOM are
-   * discouraged.
+   * UTF-8 BOM, the parser skips it.
    *
    * You do NOT need to pre-allocate a parser.  This function takes care of
    * pre-allocating a capacity defined by the batch_size defined when creating the

include/simdjson/generic/ondemand/parser-inl.h

@@ -46,6 +46,8 @@ simdjson_warn_unused simdjson_inline error_code parser::allocate(size_t new_capa
 simdjson_warn_unused simdjson_inline simdjson_result<document> parser::iterate(padded_string_view json) & noexcept {
   if (json.padding() < SIMDJSON_PADDING) { return INSUFFICIENT_PADDING; }
 
+  json.remove_utf8_bom();
+
   // Allocate if needed
   if (capacity() < json.length() || !string_buf) {
     SIMDJSON_TRY( allocate(json.length(), max_depth()) );
@@ -96,6 +98,8 @@ simdjson_warn_unused simdjson_inline simdjson_result<document> parser::iterate(c
 simdjson_warn_unused simdjson_inline simdjson_result<json_iterator> parser::iterate_raw(padded_string_view json) & noexcept {
   if (json.padding() < SIMDJSON_PADDING) { return INSUFFICIENT_PADDING; }
 
+  json.remove_utf8_bom();
+
   // Allocate if needed
   if (capacity() < json.length()) {
     SIMDJSON_TRY( allocate(json.length(), max_depth()) );
@@ -108,6 +112,10 @@ simdjson_warn_unused simdjson_inline simdjson_result<json_iterator> parser::iter
 
 inline simdjson_result<document_stream> parser::iterate_many(const uint8_t *buf, size_t len, size_t batch_size, bool allow_comma_separated) noexcept {
   if(batch_size < MINIMAL_BATCH_SIZE) { batch_size = MINIMAL_BATCH_SIZE; }
+  if((len >= 3) && (std::memcmp(buf, "\xEF\xBB\xBF", 3) == 0)) {
+    buf += 3;
+    len -= 3;
+  }
   if(allow_comma_separated && batch_size < len) { batch_size = len; }
   return document_stream(*this, buf, len, batch_size, allow_comma_separated);
 }

include/simdjson/generic/ondemand/parser.h

@@ -59,7 +59,7 @@ class parser {
    * It is expected that the content is a valid UTF-8 file, containing a valid JSON document.
    * Otherwise the iterate method may return an error. In particular, the whole input should be
    * valid: we do not attempt to tolerate incorrect content either before or after a JSON
-   * document.
+   * document. If there is a UTF-8 BOM, the parser skips it.
    *
    * ### IMPORTANT: Validate what you use
    *
@@ -188,6 +188,7 @@ class parser {
    * arrays or objects) MUST be separated with ASCII whitespace.
    *
    * The characters inside a JSON document, and between JSON documents, must be valid Unicode (UTF-8).
+   * If there is a UTF-8 BOM, the parser skips it.
    *
    * The documents must not exceed batch_size bytes (by default 1MB) or they will fail to parse.
    * Setting batch_size to excessively large or excessively small values may impact negatively the

include/simdjson/padded_string_view-inl.h

@@ -2,9 +2,10 @@
 #define SIMDJSON_PADDED_STRING_VIEW_INL_H
 
 #include "simdjson/padded_string_view.h"
-
 #include "simdjson/error-inl.h"
 
+#include <cstring> /* memcmp */
+
 namespace simdjson {
 
 inline padded_string_view::padded_string_view(const char* s, size_t len, size_t capacity) noexcept
@@ -31,6 +32,16 @@ inline size_t padded_string_view::capacity() const noexcept { return _capacity;
 
 inline size_t padded_string_view::padding() const noexcept { return capacity() - length(); }
 
+inline bool padded_string_view::remove_utf8_bom() noexcept {
+  if(length() < 3) { return false; }
+  if (std::memcmp(data(), "\xEF\xBB\xBF", 3) == 0) {
+    remove_prefix(3);
+    _capacity -= 3;
+    return true;
+  }
+  return false;
+}
+
 #if SIMDJSON_EXCEPTIONS
 inline std::ostream& operator<<(std::ostream& out, simdjson_result<padded_string_view> &s) noexcept(false) { return out << s.value(); }
 #endif

include/simdjson/padded_string_view.h

@@ -54,6 +54,13 @@ class padded_string_view : public std::string_view {
   /** The number of allocated bytes. */
   inline size_t capacity() const noexcept;
 
+  /**
+   * Remove the UTF-8 Byte Order Mark (BOM) if it exists.
+   *
+   * @return whether a BOM was found and removed
+   */
+  inline bool remove_utf8_bom() noexcept;
+
   /** The amount of padding on the string (capacity() - length()) */
   inline size_t padding() const noexcept;
 

tests/dom/basictests.cpp

@@ -66,6 +66,16 @@ namespace number_tests {
     return true;
   }
 
+  bool bomskip() {
+    TEST_START();
+    simdjson::dom::parser parser;
+    simdjson::padded_string docdata = "\xEF\xBB\xBF{\"score\":0.8825149536132812}"_padded;
+    double score;
+    ASSERT_SUCCESS(parser.parse(docdata)["score"].get_double().get(score));
+    ASSERT_EQUAL(score, 0.8825149536132812);
+    TEST_SUCCEED();
+  }
+
   bool issue2017() {
     TEST_START();
     simdjson::dom::parser parser;
@@ -386,7 +396,8 @@ namespace number_tests {
   }
 
   bool run() {
-    return issue2017() &&
+    return bomskip() &&
+           issue2017() &&
            truncated_borderline() &&
            specific_tests() &&
            ground_truth() &&
@@ -1658,6 +1669,7 @@ namespace validate_tests {
     }
     return true;
   }
+
   bool test_validate() {
     std::cout << "Running " << __func__ << std::endl;
     const std::string test = R"({ "foo" : 1, "bar" : [ 1, 2, 3 ], "baz": { "a": 1, "b": 2, "c": 3 } })";
@@ -1666,6 +1678,7 @@ namespace validate_tests {
     }
     return true;
   }
+
   bool test_range() {
     std::cout << "Running " << __func__ << std::endl;
     for(size_t len = 0; len <= 128; len++) {
@@ -1683,6 +1696,7 @@ namespace validate_tests {
     }
     return true;
   }
+
   bool test_issue1169() {
     std::cout << "Running " << __func__ << std::endl;
     std::vector<uint8_t> source(64,' ');
@@ -1693,6 +1707,7 @@ namespace validate_tests {
     }
     return true;
   }
+
   bool test_issue1169_long() {
     std::cout << "Running " << __func__ << std::endl;
     for(size_t len = 1; len <= 128; len++) {
@@ -1702,6 +1717,7 @@ namespace validate_tests {
     }
     return true;
   }
+
   bool test_random() {
     std::cout << "Running " << __func__ << std::endl;
     std::vector<uint8_t> source(64,' ');
@@ -1763,6 +1779,7 @@ namespace minify_tests {
     }
     return true;
   }
+
   // this is meant to test buffer overflows.
   bool test_various_lengths2() {
     std::cout << "Running " << __func__ << std::endl;
@@ -1781,6 +1798,7 @@ namespace minify_tests {
     }
     return true;
   }
+
   bool test_single_quote() {
     std::cout << "Running " << __func__ << std::endl;
     const std::string test = "\"";
@@ -1801,12 +1819,14 @@ namespace minify_tests {
     const std::string minified(R"({"foo":1,"bar":[1,2,0.11111111111111113],"baz":{"a":3.1415926535897936,"b":2,"c":3.141592653589794}})");
     return check_minification(test.c_str(), test.size(), minified.c_str(), minified.size());
   }
+
   bool test_minify_array() {
     std::cout << "Running " << __func__ << std::endl;
     std::string test("[ 1,    2,    3]");
     std::string minified("[1,2,3]");
     return check_minification(test.c_str(), test.size(), minified.c_str(), minified.size());
   }
+
   bool test_minify_object() {
     std::cout << "Running " << __func__ << std::endl;
     std::string test(R"({ "foo   " : 1, "b  ar" : [ 1, 2, 3 ], "baz": { "a": 1, "b": 2, "c": 3 } })");
@@ -1849,6 +1869,7 @@ namespace format_tests {
     s << doc;
     return assert_minified(s);
   }
+
   bool print_minify_parser_parse() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1868,6 +1889,7 @@ namespace format_tests {
     s << value;
     return assert_minified(s, "1");
   }
+
   bool print_minify_element() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1887,6 +1909,7 @@ namespace format_tests {
     s << array;
     return assert_minified(s, "[1,2,0.11111111111111113]");
   }
+
   bool print_minify_array() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1906,6 +1929,7 @@ namespace format_tests {
     s << object;
     return assert_minified(s, R"({"a":3.1415926535897936,"b":2,"c":3.141592653589794})");
   }
+
   bool print_minify_object() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1925,6 +1949,7 @@ namespace format_tests {
     s << parser.parse(DOCUMENT);
     return assert_minified(s);
   }
+
   bool print_minify_parser_parse_exception() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1940,6 +1965,7 @@ namespace format_tests {
     s << parser.parse(DOCUMENT)["foo"];
     return assert_minified(s, "1");
   }
+
   bool print_minify_element_result_exception() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1956,6 +1982,7 @@ namespace format_tests {
     s << value;
     return assert_minified(s, "1");
   }
+
   bool print_minify_element_exception() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;
@@ -1972,6 +1999,7 @@ namespace format_tests {
     s << parser.parse(DOCUMENT)["bar"].get_array();
     return assert_minified(s, "[1,2,0.11111111111111113]");
   }
+
   bool print_minify_array_result_exception() {
     std::cout << "Running " << __func__ << std::endl;
     dom::parser parser;

tests/dom/document_stream_tests.cpp

@@ -397,6 +397,33 @@ namespace document_stream_tests {
     std::cout << "number of documents " << count << std::endl;
     return count == 1;
   }
+
+  bool skipbom() {
+    std::cout << "Running " << __func__ << std::endl;
+    simdjson::dom::parser parser;
+    auto json = "\xEF\xBB\xBF{\"hello\": \"world\"}"_padded;
+    simdjson::dom::document_stream stream;
+    ASSERT_SUCCESS(parser.parse_many(json).get(stream));
+    size_t count = 0;
+    for (auto doc : stream) {
+        if(doc.error()) {
+          std::cerr << "Unexpected error: " << doc.error() << std::endl;
+          return false;
+        }
+        std::string expected = R"({"hello":"world"})";
+        simdjson::dom::element this_document;
+        ASSERT_SUCCESS(doc.get(this_document));
+
+        std::string answer = simdjson::minify(this_document);
+        if(answer != expected) {
+          std::cout << this_document << std::endl;
+          return false;
+        }
+        count += 1;
+    }
+    std::cout << "number of documents " << count << std::endl;
+    return count == 1;
+  }
 #if SIMDJSON_EXCEPTIONS
   bool single_document_exceptions() {
     std::cout << "Running " << __func__ << std::endl;
@@ -913,7 +940,8 @@ namespace document_stream_tests {
   }
 
   bool run() {
-    return fuzzaccess() &&
+    return skipbom() &&
+           fuzzaccess() &&
            baby_fuzzer() &&
            issue1649() &&
            adversarial_single_document_array() &&