aboutsummaryrefslogtreecommitdiff
path: root/Marlin/src/libs/numtostr.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'Marlin/src/libs/numtostr.cpp')
-rw-r--r--Marlin/src/libs/numtostr.cpp408
1 files changed, 408 insertions, 0 deletions
diff --git a/Marlin/src/libs/numtostr.cpp b/Marlin/src/libs/numtostr.cpp
new file mode 100644
index 0000000..90696e9
--- /dev/null
+++ b/Marlin/src/libs/numtostr.cpp
@@ -0,0 +1,408 @@
+/**
+ * Marlin 3D Printer Firmware
+ * Copyright (c) 2020 MarlinFirmware [https://github.com/MarlinFirmware/Marlin]
+ *
+ * Based on Sprinter and grbl.
+ * Copyright (c) 2011 Camiel Gubbels / Erik van der Zalm
+ *
+ * This program is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ *
+ */
+
+#include "numtostr.h"
+
+#include "../inc/MarlinConfigPre.h"
+#include "../core/utility.h"
+
+char conv[8] = { 0 };
+
+#define DIGIT(n) ('0' + (n))
+#define DIGIMOD(n, f) DIGIT((n)/(f) % 10)
+#define RJDIGIT(n, f) ((n) >= (f) ? DIGIMOD(n, f) : ' ')
+#define MINUSOR(n, alt) (n >= 0 ? (alt) : (n = -n, '-'))
+#define INTFLOAT(V,N) (((V) * 10 * pow(10, N) + ((V) < 0 ? -5: 5)) / 10) // pow10?
+#define UINTFLOAT(V,N) INTFLOAT((V) < 0 ? -(V) : (V), N)
+
+// Format uint8_t (0-100) as rj string with 123% / _12% / __1% format
+const char* pcttostrpctrj(const uint8_t i) {
+ conv[3] = RJDIGIT(i, 100);
+ conv[4] = RJDIGIT(i, 10);
+ conv[5] = DIGIMOD(i, 1);
+ conv[6] = '%';
+ return &conv[3];
+}
+
+// Convert uint8_t (0-255) to a percentage, format as above
+const char* ui8tostr4pctrj(const uint8_t i) {
+ return pcttostrpctrj(ui8_to_percent(i));
+}
+
+// Convert unsigned 8bit int to string 123 format
+const char* ui8tostr3rj(const uint8_t i) {
+ conv[4] = RJDIGIT(i, 100);
+ conv[5] = RJDIGIT(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[4];
+}
+
+// Convert uint8_t to string with 12 format
+const char* ui8tostr2(const uint8_t i) {
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[5];
+}
+
+// Convert signed 8bit int to rj string with 123 or -12 format
+const char* i8tostr3rj(const int8_t x) {
+ int xx = x;
+ conv[4] = MINUSOR(xx, RJDIGIT(xx, 100));
+ conv[5] = RJDIGIT(xx, 10);
+ conv[6] = DIGIMOD(xx, 1);
+ return &conv[4];
+}
+
+#if HAS_PRINT_PROGRESS_PERMYRIAD
+ // Convert unsigned 16-bit permyriad to percent with 100 / 23 / 23.4 / 3.45 format
+ const char* permyriadtostr4(const uint16_t xx) {
+ if (xx >= 10000)
+ return "100";
+ else if (xx >= 1000) {
+ conv[3] = DIGIMOD(xx, 1000);
+ conv[4] = DIGIMOD(xx, 100);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(xx, 10);
+ return &conv[3];
+ }
+ else if (xx % 100 == 0) {
+ conv[4] = ' ';
+ conv[5] = RJDIGIT(xx, 1000);
+ conv[6] = DIGIMOD(xx, 100);
+ return &conv[4];
+ }
+ else {
+ conv[3] = DIGIMOD(xx, 100);
+ conv[4] = '.';
+ conv[5] = DIGIMOD(xx, 10);
+ conv[6] = RJDIGIT(xx, 1);
+ return &conv[3];
+ }
+ }
+#endif
+
+// Convert unsigned 16bit int to string 12345 format
+const char* ui16tostr5rj(const uint16_t xx) {
+ conv[2] = RJDIGIT(xx, 10000);
+ conv[3] = RJDIGIT(xx, 1000);
+ conv[4] = RJDIGIT(xx, 100);
+ conv[5] = RJDIGIT(xx, 10);
+ conv[6] = DIGIMOD(xx, 1);
+ return &conv[2];
+}
+
+// Convert unsigned 16bit int to string 1234 format
+const char* ui16tostr4rj(const uint16_t xx) {
+ conv[3] = RJDIGIT(xx, 1000);
+ conv[4] = RJDIGIT(xx, 100);
+ conv[5] = RJDIGIT(xx, 10);
+ conv[6] = DIGIMOD(xx, 1);
+ return &conv[3];
+}
+
+// Convert unsigned 16bit int to string 123 format
+const char* ui16tostr3rj(const uint16_t xx) {
+ conv[4] = RJDIGIT(xx, 100);
+ conv[5] = RJDIGIT(xx, 10);
+ conv[6] = DIGIMOD(xx, 1);
+ return &conv[4];
+}
+
+// Convert signed 16bit int to rj string with 123 or -12 format
+const char* i16tostr3rj(const int16_t x) {
+ int xx = x;
+ conv[4] = MINUSOR(xx, RJDIGIT(xx, 100));
+ conv[5] = RJDIGIT(xx, 10);
+ conv[6] = DIGIMOD(xx, 1);
+ return &conv[4];
+}
+
+// Convert unsigned 16bit int to lj string with 123 format
+const char* i16tostr3left(const int16_t i) {
+ char *str = &conv[6];
+ *str = DIGIMOD(i, 1);
+ if (i >= 10) {
+ *(--str) = DIGIMOD(i, 10);
+ if (i >= 100)
+ *(--str) = DIGIMOD(i, 100);
+ }
+ return str;
+}
+
+// Convert signed 16bit int to rj string with 1234, _123, -123, _-12, or __-1 format
+const char* i16tostr4signrj(const int16_t i) {
+ const bool neg = i < 0;
+ const int ii = neg ? -i : i;
+ if (i >= 1000) {
+ conv[3] = DIGIMOD(ii, 1000);
+ conv[4] = DIGIMOD(ii, 100);
+ conv[5] = DIGIMOD(ii, 10);
+ }
+ else if (ii >= 100) {
+ conv[3] = neg ? '-' : ' ';
+ conv[4] = DIGIMOD(ii, 100);
+ conv[5] = DIGIMOD(ii, 10);
+ }
+ else {
+ conv[3] = ' ';
+ conv[4] = ' ';
+ if (ii >= 10) {
+ conv[4] = neg ? '-' : ' ';
+ conv[5] = DIGIMOD(ii, 10);
+ }
+ else {
+ conv[5] = neg ? '-' : ' ';
+ }
+ }
+ conv[6] = DIGIMOD(ii, 1);
+ return &conv[3];
+}
+
+// Convert unsigned float to string with 1.23 format
+const char* ftostr12ns(const float &f) {
+ const long i = UINTFLOAT(f, 2);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = '.';
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[3];
+}
+
+// Convert unsigned float to string with 12.3 format
+const char* ftostr31ns(const float &f) {
+ const long i = UINTFLOAT(f, 1);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = DIGIMOD(i, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[3];
+}
+
+// Convert unsigned float to string with 123.4 format
+const char* ftostr41ns(const float &f) {
+ const long i = UINTFLOAT(f, 1);
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = DIGIMOD(i, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[2];
+}
+
+// Convert signed float to fixed-length string with 12.34 / _2.34 / -2.34 or -23.45 / 123.45 format
+const char* ftostr42_52(const float &f) {
+ if (f <= -10 || f >= 100) return ftostr52(f); // -23.45 / 123.45
+ long i = INTFLOAT(f, 2);
+ conv[2] = (f >= 0 && f < 10) ? ' ' : MINUSOR(i, DIGIMOD(i, 1000));
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = '.';
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[2];
+}
+
+// Convert signed float to fixed-length string with 023.45 / -23.45 format
+const char* ftostr52(const float &f) {
+ long i = INTFLOAT(f, 2);
+ conv[1] = MINUSOR(i, DIGIMOD(i, 10000));
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = '.';
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[1];
+}
+
+// Convert signed float to fixed-length string with 12.345 / _2.345 / -2.345 or -23.45 / 123.45 format
+const char* ftostr53_63(const float &f) {
+ if (f <= -10 || f >= 100) return ftostr63(f); // -23.456 / 123.456
+ long i = INTFLOAT(f, 3);
+ conv[1] = (f >= 0 && f < 10) ? ' ' : MINUSOR(i, DIGIMOD(i, 10000));
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = '.';
+ conv[4] = DIGIMOD(i, 100);
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[1];
+}
+
+// Convert signed float to fixed-length string with 023.456 / -23.456 format
+const char* ftostr63(const float &f) {
+ long i = INTFLOAT(f, 3);
+ conv[0] = MINUSOR(i, DIGIMOD(i, 100000));
+ conv[1] = DIGIMOD(i, 10000);
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = '.';
+ conv[4] = DIGIMOD(i, 100);
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[0];
+}
+
+#if ENABLED(LCD_DECIMAL_SMALL_XY)
+
+ // Convert float to rj string with 1234, _123, -123, _-12, 12.3, _1.2, or -1.2 format
+ const char* ftostr4sign(const float &f) {
+ const int i = INTFLOAT(f, 1);
+ if (!WITHIN(i, -99, 999)) return i16tostr4signrj((int)f);
+ const bool neg = i < 0;
+ const int ii = neg ? -i : i;
+ conv[3] = neg ? '-' : (ii >= 100 ? DIGIMOD(ii, 100) : ' ');
+ conv[4] = DIGIMOD(ii, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(ii, 1);
+ return &conv[3];
+ }
+
+#endif
+
+// Convert float to fixed-length string with +12.3 / -12.3 format
+const char* ftostr31sign(const float &f) {
+ int i = INTFLOAT(f, 1);
+ conv[2] = MINUSOR(i, '+');
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = DIGIMOD(i, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[2];
+}
+
+// Convert float to fixed-length string with +123.4 / -123.4 format
+const char* ftostr41sign(const float &f) {
+ int i = INTFLOAT(f, 1);
+ conv[1] = MINUSOR(i, '+');
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = DIGIMOD(i, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[1];
+}
+
+// Convert signed float to string (6 digit) with -1.234 / _0.000 / +1.234 format
+const char* ftostr43sign(const float &f, char plus/*=' '*/) {
+ long i = INTFLOAT(f, 3);
+ conv[1] = i ? MINUSOR(i, plus) : ' ';
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = '.';
+ conv[4] = DIGIMOD(i, 100);
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[1];
+}
+
+// Convert signed float to string (5 digit) with -1.2345 / _0.0000 / +1.2345 format
+const char* ftostr54sign(const float &f, char plus/*=' '*/) {
+ long i = INTFLOAT(f, 4);
+ conv[0] = i ? MINUSOR(i, plus) : ' ';
+ conv[1] = DIGIMOD(i, 10000);
+ conv[2] = '.';
+ conv[3] = DIGIMOD(i, 1000);
+ conv[4] = DIGIMOD(i, 100);
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return &conv[0];
+}
+
+// Convert unsigned float to rj string with 12345 format
+const char* ftostr5rj(const float &f) {
+ const long i = UINTFLOAT(f, 0);
+ return ui16tostr5rj(i);
+}
+
+// Convert signed float to string with +1234.5 format
+const char* ftostr51sign(const float &f) {
+ long i = INTFLOAT(f, 1);
+ conv[0] = MINUSOR(i, '+');
+ conv[1] = DIGIMOD(i, 10000);
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = DIGIMOD(i, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(i, 1);
+ return conv;
+}
+
+// Convert signed float to string with +123.45 format
+const char* ftostr52sign(const float &f) {
+ long i = INTFLOAT(f, 2);
+ conv[0] = MINUSOR(i, '+');
+ conv[1] = DIGIMOD(i, 10000);
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = DIGIMOD(i, 100);
+ conv[4] = '.';
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return conv;
+}
+
+// Convert signed float to string with +12.345 format
+const char* ftostr53sign(const float &f) {
+ long i = INTFLOAT(f, 3);
+ conv[0] = MINUSOR(i, '+');
+ conv[1] = DIGIMOD(i, 10000);
+ conv[2] = DIGIMOD(i, 1000);
+ conv[3] = '.';
+ conv[4] = DIGIMOD(i, 100);
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIMOD(i, 1);
+ return conv;
+}
+
+// Convert unsigned float to string with ____4.5, __34.5, _234.5, 1234.5 format
+const char* ftostr51rj(const float &f) {
+ const long i = UINTFLOAT(f, 1);
+ conv[0] = ' ';
+ conv[1] = RJDIGIT(i, 10000);
+ conv[2] = RJDIGIT(i, 1000);
+ conv[3] = RJDIGIT(i, 100);
+ conv[4] = DIGIMOD(i, 10);
+ conv[5] = '.';
+ conv[6] = DIGIMOD(i, 1);
+ return conv;
+}
+
+// Convert signed float to space-padded string with -_23.4_ format
+const char* ftostr52sp(const float &f) {
+ long i = INTFLOAT(f, 2);
+ uint8_t dig;
+ conv[0] = MINUSOR(i, ' ');
+ conv[1] = RJDIGIT(i, 10000);
+ conv[2] = RJDIGIT(i, 1000);
+ conv[3] = DIGIMOD(i, 100);
+
+ if ((dig = i % 10)) { // second digit after decimal point?
+ conv[4] = '.';
+ conv[5] = DIGIMOD(i, 10);
+ conv[6] = DIGIT(dig);
+ }
+ else {
+ if ((dig = (i / 10) % 10)) { // first digit after decimal point?
+ conv[4] = '.';
+ conv[5] = DIGIT(dig);
+ }
+ else // nothing after decimal point
+ conv[4] = conv[5] = ' ';
+ conv[6] = ' ';
+ }
+ return conv;
+}