- introducing spdlog / libdiagnostics based logging/diagnostics outpu…

…t over tprintf*() functions.

- remove experimental LogLevelElevation code.
- make sure the debug_all preset is set up BEFORE any command-line arguments direct tesseract to set some arbitrary parameters just below, for otherwise those `-c xyz=v` commands may be overruled by the debug_all preset!
- clean up the CLI options/commands parsing & processing: detect and process the various commands more obviously and use an enum/int variable to encode the active command to be executed by tesseract.

src/ccmain/tessedit.cpp

@@ -302,6 +302,55 @@ void Tesseract::ParseLanguageString(const std::string &lang_str, std::vector<std
   }
 }
 
+// Parse a string of the form `<box>[+<box>]*` where box is given as
+// `lNtNwNhN` or `lNtNrNbN` with the `N` being numeric values.
+//
+// Returns an BOXA instance (array of BOX coordinates) on success or NULL on failure.
+// Errors are reported via tprintError() as they happen.
+BOXA *Tesseract::ParseRectsString(const char *rects_str) {
+  // Dev Note: use classic C code approach instead of C++ std::string based: much easier & less heap thrashing.
+  char *rects = strdup(rects_str);
+
+  // also match ',' and ';', as well as '+', in case user used one of those separators instead of '+':
+  BOXA *boxa = boxaCreate(100);
+  int idx = 0;
+  char *token = rects;
+  for (;;) {
+    int pos = strspn(token, " :;+");
+    token += pos;
+    pos = strcspn(rects, " :;+");
+    bool eol = (token[pos] == 0);
+    token[pos] = 0;
+
+    // as an extra service, convert to lowercase before parsing:
+    strlwr(token);
+
+    int left, top, width, height, right, bottom;
+    int params = sscanf(token, "l%dt%dw%dh%d", &left, &top, &width, &height);
+    if (params == 4) {
+      BOX *box = boxCreateValid(left, top, width, height);
+      boxaAddBox(boxa, box, L_INSERT);
+    } else {
+      params = sscanf(token, "l%dt%dr%db%d", &left, &top, &right, &bottom);
+      if (params == 4) {
+        BOX *box = boxCreateValid(left, top, right - left, bottom - top);
+        boxaAddBox(boxa, box, L_INSERT);
+      } else {
+        tprintError("Rectangle spec line part '{}' does not match either of the supported formats LTDH or LTRB, f.e. something akin to 'l30t60w50h100'. Your line:\n    {}\n", token, rects_str);
+        boxaDestroy(&boxa);
+        return nullptr;
+      }
+    }
+    token += pos;
+    if (eol) {
+      break;
+    }
+    token++;
+  }
+  return boxa;
+}
+
+
 // Initialize for potentially a set of languages defined by the language
 // string and recursively any additional languages required by any language
 // traineddata file (via tessedit_load_sublangs in its config) that is loaded.

src/ccutil/helpers.h

@@ -133,7 +133,7 @@ inline void chomp_string(char *str) {
 }
 
 // return the smallest multiple of block_size greater than or equal to n.
-inline int RoundUp(int n, int block_size) {
+static inline int RoundUp(int n, int block_size) {
   return block_size * ((n + block_size - 1) / block_size);
 }
 
@@ -187,7 +187,7 @@ inline void IntersectRange(const T &lower1, const T &upper1, T *lower2, T *upper
 // Proper modulo arithmetic operator. Returns a mod b that works for -ve a.
 // For any integer a and positive b, returns r : 0<=r<b and a=n*b + r for
 // some integer n.
-inline int Modulo(int a, int b) {
+static inline int Modulo(int a, int b) {
   return (a % b + b) % b;
 }
 
@@ -196,29 +196,29 @@ inline int Modulo(int a, int b) {
 // counting at 0. With simple rounding 1/3 = 0, 0/3 = 0 -1/3 = 0, -2/3 = 0,
 // -3/3 = 0 and -4/3 = -1.
 // I want 1/3 = 0, 0/3 = 0, -1/3 = 0, -2/3 = -1, -3/3 = -1 and -4/3 = -1.
-inline int DivRounded(int a, int b) {
+static inline int DivRounded(int a, int b) {
   if (b < 0) {
     return -DivRounded(a, -b);
   }
   return a >= 0 ? (a + b / 2) / b : (a - b / 2) / b;
 }
 
 // Return a double cast to int with rounding.
-inline int IntCastRounded(double x) {
+static inline int IntCastRounded(double x) {
   ASSERT0(std::isfinite(x));
   ASSERT0(x < INT_MAX);
   ASSERT0(x > INT_MIN);
   return x >= 0.0 ? static_cast<int>(x + 0.5) : -static_cast<int>(-x + 0.5);
 }
 
 // Return a float cast to int with rounding.
-inline int IntCastRounded(float x) {
+static inline int IntCastRounded(float x) {
   ASSERT0(std::isfinite(x));
   return x >= 0.0F ? static_cast<int>(x + 0.5F) : -static_cast<int>(-x + 0.5F);
 }
 
 // Reverse the order of bytes in a n byte quantity for big/little-endian switch.
-inline void ReverseN(void *ptr, int num_bytes) {
+static inline void ReverseN(void *ptr, int num_bytes) {
   ASSERT0(num_bytes == 1 || num_bytes == 2 || num_bytes == 4 || num_bytes == 8);
   char *cptr = static_cast<char *>(ptr);
   int halfsize = num_bytes / 2;
@@ -230,7 +230,7 @@ inline void ReverseN(void *ptr, int num_bytes) {
 }
 
 // Reverse the order of bytes in a 32 bit quantity for big/little-endian switch.
-inline void Reverse32(void *ptr) {
+static inline void Reverse32(void *ptr) {
   ReverseN(ptr, 4);
 }
 

src/ccutil/pathutils.h

@@ -15,7 +15,7 @@ namespace fs {
   using namespace ::std::filesystem;
 #endif
 
-  bool exists(const char *filename) {
+  static inline bool exists(const char *filename) {
 #if defined(WIN32) || defined(_WIN32) || defined(_WIN64)
     return _access(filename, 0) == 0;
 #else

src/ccutil/tprintf.cpp

@@ -28,6 +28,8 @@
 #include <climits> // for INT_MAX
 #include <cstdio>
 
+#include <diagnostics/diagnostics.h>
+
 #ifdef HAVE_MUPDF
 
 #include "mupdf/fitz/config.h"
@@ -40,62 +42,81 @@
 
 namespace tesseract {
 
-#ifdef HAVE_MUPDF
+static void assert_that_a_spdlog_sink_and_logger_are_active() {
+  std::shared_ptr<spdlog::logger> logger = spdlog::default_logger();
+  ASSERT0(!!logger);
+  const std::vector<spdlog::sink_ptr> &sinks = logger->sinks();
+  ASSERT0(!sinks.empty());
+}
+
 
 // Warning: tprintf() is invoked in tesseract for PARTIAL lines, so we SHOULD gather these fragments
 // here before dispatching the gathered lines to the appropriate back-end API!
-static void fz_tess_tprintf(int level, fmt::string_view format, fmt::format_args args) {
-  static int block_level = T_LOG_DEBUG;
+static void gather_and_log_a_single_tprintf_line(int level, fmt::string_view format, fmt::format_args args) {
+  static int block_level = T_LOG_TRACE;
+
+  // elevation means LOWERING the level value as lower is higher severity!
+  level -= tprintGetLevelElevation();
+
   // sanity check/clipping: there's no log level beyond ERROR severity: ERROR is the highest it can possibly get.
   if (level < T_LOG_ERROR) {
-	  level = T_LOG_ERROR;
-  }
-  // make the entire message line have the most severe log level given for any part of the line:
-  if (level < block_level) {
-    block_level = level;
+    level = T_LOG_ERROR;
   }
 
   auto msg = fmt::vformat(format, args);
 
+  // when this is a partial message, store it in the buffer until later, when the message is completed.
   static std::string msg_buffer;
-  msg_buffer += msg;
-  if (!msg_buffer.ends_with('\n'))
+  if (!msg.ends_with('\n')) {
+    // make the entire message line have the most severe log level given for any part of the line:
+    if (level < block_level) {
+      block_level = level;
+    }
+    msg_buffer += msg;
     return;
+  }
 
-  const char *s = msg_buffer.c_str();
-  level = block_level;
-
-  if (!strncmp(s, "ERROR: ", 7))
-    fz_error(NULL, "%s", s + 7);
-  else if (!strncmp(s, "WARNING: ", 9))
-    fz_warn(NULL, "%s", s + 9);
-  else {
-	switch (level) {
-	case T_LOG_ERROR:
-      fz_error(NULL, "%s", s);
-	  break;
-	case T_LOG_WARN:
-	  fz_warn(NULL, "%s", s);
-	  break;
-	case T_LOG_INFO:
-      fz_info(NULL, "%s", s);
-	  break;
-	case T_LOG_DEBUG:
-	default:
-      fz_info(NULL, "%s", s);
-	  break;
-	}
+  // `msg` carries a complete message, or at least the end of it:
+  // when there's some old stuff waiting for us: append and pick up the tracked error level.
+  if (!msg_buffer.empty()) {
+    level = block_level;
+    msg = msg_buffer + msg;
+    msg_buffer.clear();
   }
 
-  msg_buffer.clear();
+  // We've gathered a single, entire, message: now output it line-by-line (if it's multi-line internally).
+  const char *s = msg.c_str();
+
+  if (!strncmp(s, "ERROR: ", 7)) {
+    s += 7;
+    if (level > T_LOG_ERROR)
+      level = T_LOG_ERROR;
+  } else if (!strncmp(s, "WARNING: ", 9)) {
+    s += 9;
+    if (level > T_LOG_WARN)
+      level = T_LOG_WARN;
+  }
+
+  switch (level) {
+    case T_LOG_ERROR:
+      spdlog::error(s);
+      break;
+    case T_LOG_WARN:
+      spdlog::warn(s);
+      break;
+    case T_LOG_INFO:
+      spdlog::info(s);
+      break;
+    case T_LOG_DEBUG:
+    default:
+      spdlog::debug(s);
+      break;
+  }
 
   // reset next line log level to lowest possible:
   block_level = T_LOG_DEBUG;
 }
 
-#endif
-
-#define MAX_MSG_LEN 2048
 
 // when we use tesseract as part of MuPDF (or mixed with it), we use the fz_error/fz_warn/fz_info APIs to
 // output any error/info/debug messages and have the callbacks which MAY be registered with those APIs
@@ -126,9 +147,10 @@ const int tprintGetLevelElevation(void)
 
 // Trace printf
 void vTessPrint(int level, fmt::string_view format, fmt::format_args args) {
-#ifdef HAVE_MUPDF
-	fz_tess_tprintf(level, format, args);
-#else
+  assert_that_a_spdlog_sink_and_logger_are_active();
+
+  gather_and_log_a_single_tprintf_line(level, format, args);
+
   const char *debug_file_name = debug_file.c_str();
   static FILE *debugfp = nullptr; // debug file
 
@@ -158,7 +180,6 @@ void vTessPrint(int level, fmt::string_view format, fmt::format_args args) {
   } else {
     fmt::vprint(stderr, format, args);
   }
-#endif
 }
 
 } // namespace tesseract

src/ccutil/tprintf.h

@@ -27,10 +27,12 @@ namespace tesseract {
 // Note: You can disable some log messages by setting FLAGS_tlog_level > 0.
 
 enum LogLevel : int {
-	T_LOG_ERROR,
+  //T_LOG_CRITICAL,
+  T_LOG_ERROR,
 	T_LOG_WARN,
 	T_LOG_INFO,
 	T_LOG_DEBUG,
+  T_LOG_TRACE,
 };
 
 // Helper function for tprintf.
@@ -44,22 +46,27 @@ extern TESS_API const int tprintGetLevelElevation(void);
 
 template <typename S, typename... Args>
 void tprintError(const S *format, Args &&...args) {
-  vTessPrint(T_LOG_ERROR + tprintGetLevelElevation(), format, fmt::make_format_args(args...));
+  vTessPrint(T_LOG_ERROR, format, fmt::make_format_args(args...));
 }
 
 template <typename S, typename... Args>
 void tprintWarn(const S *format, Args &&...args) {
-	vTessPrint(T_LOG_WARN + tprintGetLevelElevation(), format, fmt::make_format_args(args...));
+	vTessPrint(T_LOG_WARN, format, fmt::make_format_args(args...));
 }
 
 template <typename S, typename... Args>
 void tprintInfo(const S *format, Args &&...args) {
-	vTessPrint(T_LOG_INFO + tprintGetLevelElevation(), format, fmt::make_format_args(args...));
+	vTessPrint(T_LOG_INFO, format, fmt::make_format_args(args...));
 }
 
 template <typename S, typename... Args>
 void tprintDebug(const S *format, Args &&...args) {
-	vTessPrint(T_LOG_DEBUG + tprintGetLevelElevation(), format, fmt::make_format_args(args...));
+	vTessPrint(T_LOG_DEBUG, format, fmt::make_format_args(args...));
+}
+
+template <typename S, typename... Args>
+void tprintTrace(const S *format, Args &&...args) {
+  vTessPrint(T_LOG_TRACE, format, fmt::make_format_args(args...));
 }
 
 } // namespace tesseract