mnbvc188199/Warrior_EA
Watch
1
0
Fork
You've already forked Warrior_EA
 1
forked from animatedread/Warrior_EA
Code Pull requests Activity
Warrior_EA/System/StatusLabel.mqh

265 lines
12 KiB
MQL5
Raw Permalink Normal View History

refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
//+------------------------------------------------------------------+
//| StatusLabel.mqh |
//| AnimateDread |
//| https://www.mql5.com |
//+------------------------------------------------------------------+
#property copyright "AnimateDread"
#property link "https://www.mql5.com"
//+------------------------------------------------------------------+
refactor(comments): box headers to stdlib length The //| box blocks were excluded from 0b06f8e and 5efdb48 and were what remained: 160 of them ran to 10+ lines, the longest to 88. Compressed to their leading topic sentences - 5 lines for a function header, 8 for a file header - keeping the box format and the standard MQL5 name/author lines verbatim. Verified at the BYTE level this time, across every in-scope file: the list of non-comment lines is byte-identical to HEAD and braces balance. The first check compared a locale-decoded 'git show' against a UTF-8 read and flagged 25 files that had not changed at all - every BOM and every non-ASCII line mismatched. 47,696 -> 40,665 lines in scope; comment share 38% -> 26%. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-22 00:30:14 -04:00
//| Replaces Comment() for on-chart status text. Comment() draws |
//| directly onto the chart surface with no background, so status |
//| text becomes unreadable over candles. |
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
//+------------------------------------------------------------------+
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
#define STATUS_LABEL_PREFIX "WarriorStatusLine"
#define STATUS_LABEL_FONTSIZE 9
#define STATUS_LABEL_FONT "Consolas"
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
#define STATUS_LABEL_PAD_X 3
#define STATUS_LABEL_PAD_Y 1
#define STATUS_LABEL_LINE_GAP 2 // extra pixels between successive lines' backgrounds
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
#define STATUS_LABEL_RIGHT_MARGIN 20
#define STATUS_LABEL_MIN_WRAP_WIDTH 150
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
#define STATUS_LABEL_MAX_CHARS 55 // safe margin under OBJPROP_TEXT's observed ~63-char truncation
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
#define STATUS_LABEL_MAX_LINES 64
int g_statusLabelLineCount = 0;
feat: implement one-shot pattern-database backfill and enhance accuracy tracking for ensemble models
2026-08-16 21:08:41 -04:00
//--- Y (chart pixels, CORNER_LEFT_UPPER) just below the last line SetStatusLabel() drew - i.e. where
//--- something else can be placed to sit UNDER the label instead of overlapping it. Defaults to the
//--- label's own default top (see SetStatusLabel()'s y=20 default) before the label has ever drawn.
int g_statusLabelBottomY = 20;
int StatusLabelBottomY(void) { return g_statusLabelBottomY; }
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
void EnsureStatusLine(int idx)
{
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
string bg = STATUS_LABEL_PREFIX + (string)idx + "BG";
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
string txt = STATUS_LABEL_PREFIX + (string)idx + "Text";
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
if(ObjectFind(0, bg) < 0)
{
ObjectCreate(0, bg, OBJ_RECTANGLE_LABEL, 0, 0, 0);
ObjectSetInteger(0, bg, OBJPROP_CORNER, CORNER_LEFT_UPPER);
ObjectSetInteger(0, bg, OBJPROP_BGCOLOR, (color)ColorToARGB(clrBlack, 150));
ObjectSetInteger(0, bg, OBJPROP_BORDER_TYPE, BORDER_FLAT);
ObjectSetInteger(0, bg, OBJPROP_COLOR, clrNONE);
ObjectSetInteger(0, bg, OBJPROP_BACK, false);
ObjectSetInteger(0, bg, OBJPROP_SELECTABLE, false);
ObjectSetInteger(0, bg, OBJPROP_SELECTED, false);
ObjectSetInteger(0, bg, OBJPROP_HIDDEN, true);
}
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
if(ObjectFind(0, txt) < 0)
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
{
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
ObjectCreate(0, txt, OBJ_LABEL, 0, 0, 0);
ObjectSetInteger(0, txt, OBJPROP_CORNER, CORNER_LEFT_UPPER);
ObjectSetInteger(0, txt, OBJPROP_ANCHOR, ANCHOR_LEFT_UPPER);
ObjectSetInteger(0, txt, OBJPROP_FONTSIZE, STATUS_LABEL_FONTSIZE);
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
ObjectSetString(0, txt, OBJPROP_FONT, STATUS_LABEL_FONT);
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
ObjectSetInteger(0, txt, OBJPROP_COLOR, clrWhite);
ObjectSetInteger(0, txt, OBJPROP_BACK, false);
ObjectSetInteger(0, txt, OBJPROP_SELECTABLE, false);
ObjectSetInteger(0, txt, OBJPROP_SELECTED, false);
ObjectSetInteger(0, txt, OBJPROP_HIDDEN, true);
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
}
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
}
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
//+------------------------------------------------------------------+
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
//| Appends one already-short-enough chunk to outLines/outCount, |
//| force-splitting it further if it still exceeds |
//| STATUS_LABEL_MAX_CHARS (a single word longer than the cap, with |
//| no spaces to break on). |
//+------------------------------------------------------------------+
void AppendStatusChunk(string chunk, string &outLines[], int &outCount)
{
int pos = 0;
int len = StringLen(chunk);
while(len - pos > STATUS_LABEL_MAX_CHARS)
{
ArrayResize(outLines, outCount + 1);
outLines[outCount++] = StringSubstr(chunk, pos, STATUS_LABEL_MAX_CHARS);
pos += STATUS_LABEL_MAX_CHARS;
}
ArrayResize(outLines, outCount + 1);
outLines[outCount++] = StringSubstr(chunk, pos);
}
//+------------------------------------------------------------------+
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
//| Greedily breaks one logical line into as many visual lines as |
//| needed to keep each one's measured TextGetSize() width within |
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
//| maxWidth AND its character count within STATUS_LABEL_MAX_CHARS - |
//| the latter is what actually matters (see this file's header |
//| comment on ObjectSetString's ~63-char truncation), the pixel width |
//| check just wraps sooner on a narrow chart. Requires TextSetFont() |
//| to already have been called by the caller so TextGetSize() |
//| measures against the same font actually rendered. |
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
//+------------------------------------------------------------------+
void WrapLineInto(string line, int maxWidth, string &outLines[], int &outCount)
{
fix(System/StatusLabel): correct type mismatches in TextGetSize usage TextGetSize returns uint values, so dummyW, dummyH, textW, and textH changed to uint. Added explicit casts to int where used in comparisons or arithmetic to prevent signed/unsigned mismatch warnings.
2026-07-17 21:37:52 -04:00
uint dummyW = 0, dummyH = 0;
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
TextGetSize(line, dummyW, dummyH);
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
if(StringLen(line) == 0 || ((int)dummyW <= maxWidth && StringLen(line) <= STATUS_LABEL_MAX_CHARS))
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
{
ArrayResize(outLines, outCount + 1);
outLines[outCount++] = line;
return;
}
string words[];
int wordCount = StringSplit(line, ' ', words);
string current = "";
for(int w = 0; w < wordCount; w++)
{
string candidate = (StringLen(current) == 0) ? words[w] : current + " " + words[w];
fix(System/StatusLabel): correct type mismatches in TextGetSize usage TextGetSize returns uint values, so dummyW, dummyH, textW, and textH changed to uint. Added explicit casts to int where used in comparisons or arithmetic to prevent signed/unsigned mismatch warnings.
2026-07-17 21:37:52 -04:00
uint cw = 0, ch = 0;
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
TextGetSize(candidate, cw, ch);
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
bool tooWide = ((int)cw > maxWidth || StringLen(candidate) > STATUS_LABEL_MAX_CHARS);
if(tooWide && StringLen(current) > 0)
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
{
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
AppendStatusChunk(current, outLines, outCount);
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
current = words[w];
}
else
current = candidate;
}
refactor: simplify on-chart status text and remove verbose metrics The training status panel was too tall and wordy, displaying detailed recall, precision, and continual-learning OOS simulation metrics that are not needed for at-a-glance monitoring. These metrics are still tracked internally for convergence gating, but removed from the on-chart display to keep the panel compact. Also shortened label names (e.g., "[IS] Accuracy" -> "IS Acc") to save space without reducing clarity.
2026-07-17 21:49:30 -04:00
if(StringLen(current) > 0)
AppendStatusChunk(current, outLines, outCount);
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
}
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
void SetStatusLabel(string text, int x = 10, int y = 20)
{
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
//--- TextSetFont's size argument is tenths of a point when negative (matches OBJPROP_FONTSIZE's
//--- point-based units above) - must be set before any TextGetSize() call since it's global state
TextSetFont(STATUS_LABEL_FONT, -STATUS_LABEL_FONTSIZE * 10, 0);
long chartWidthPx = 0;
ChartGetInteger(0, CHART_WIDTH_IN_PIXELS, 0, chartWidthPx);
int maxWidth = (int)chartWidthPx - x - STATUS_LABEL_RIGHT_MARGIN;
if(maxWidth < STATUS_LABEL_MIN_WRAP_WIDTH)
maxWidth = STATUS_LABEL_MIN_WRAP_WIDTH;
string rawLines[];
int rawCount = StringSplit(text, '\n', rawLines);
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
string lines[];
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
int lineCount = 0;
for(int r = 0; r < rawCount && lineCount < STATUS_LABEL_MAX_LINES; r++)
WrapLineInto(rawLines[r], maxWidth, lines, lineCount);
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
if(lineCount > STATUS_LABEL_MAX_LINES)
lineCount = STATUS_LABEL_MAX_LINES;
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
int lineY = y;
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
for(int i = 0; i < lineCount; i++)
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
{
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
EnsureStatusLine(i);
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
string bg = STATUS_LABEL_PREFIX + (string)i + "BG";
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
string txt = STATUS_LABEL_PREFIX + (string)i + "Text";
fix(System/StatusLabel): correct type mismatches in TextGetSize usage TextGetSize returns uint values, so dummyW, dummyH, textW, and textH changed to uint. Added explicit casts to int where used in comparisons or arithmetic to prevent signed/unsigned mismatch warnings.
2026-07-17 21:37:52 -04:00
uint textW = 0, textH = 0;
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
//--- an empty line measures 0x0, which would collapse its background - fall back to a single
//--- space so blank separator lines still reserve a row of height
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
TextGetSize(StringLen(lines[i]) > 0 ? lines[i] : " ", textW, textH);
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
ObjectSetInteger(0, bg, OBJPROP_XDISTANCE, x - STATUS_LABEL_PAD_X);
ObjectSetInteger(0, bg, OBJPROP_YDISTANCE, lineY - STATUS_LABEL_PAD_Y);
ObjectSetInteger(0, bg, OBJPROP_XSIZE, (int)textW + STATUS_LABEL_PAD_X * 2);
ObjectSetInteger(0, bg, OBJPROP_YSIZE, (int)textH + STATUS_LABEL_PAD_Y * 2);
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
ObjectSetInteger(0, txt, OBJPROP_XDISTANCE, x);
ObjectSetInteger(0, txt, OBJPROP_YDISTANCE, lineY);
ObjectSetString(0, txt, OBJPROP_TEXT, lines[i]);
fix(System/StatusLabel): correct type mismatches in TextGetSize usage TextGetSize returns uint values, so dummyW, dummyH, textW, and textH changed to uint. Added explicit casts to int where used in comparisons or arithmetic to prevent signed/unsigned mismatch warnings.
2026-07-17 21:37:52 -04:00
lineY += (int)textH + STATUS_LABEL_LINE_GAP;
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
}
feat: implement one-shot pattern-database backfill and enhance accuracy tracking for ensemble models
2026-08-16 21:08:41 -04:00
g_statusLabelBottomY = lineY;
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
//--- fewer lines than last call - remove the now-unused trailing line objects rather than
//--- leaving stale text (e.g. a previous longer status block) hanging below the new one
for(int i = lineCount; i < g_statusLabelLineCount; i++)
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
{
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
string bg = STATUS_LABEL_PREFIX + (string)i + "BG";
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
string txt = STATUS_LABEL_PREFIX + (string)i + "Text";
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
if(ObjectFind(0, bg) >= 0)
ObjectDelete(0, bg);
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
if(ObjectFind(0, txt) >= 0)
ObjectDelete(0, txt);
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
}
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
g_statusLabelLineCount = lineCount;
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
//--- ObjectSet*() alone does not trigger a repaint - without this the new text/background only
//--- appears once something else forces a redraw (dragging the chart, clicking an object, a new
//--- bar), which made updates look stalled between those events
ChartRedraw(0);
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
}
//+------------------------------------------------------------------+
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
//| Removes every status line object. Mirrors the old Comment("") |
//| clear on shutdown - ObjectsDeleteAll(0) elsewhere clears chart |
//| objects, but callers still need an explicit clear at points where |
//| only the status text (not the whole chart) should be reset. |
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
//+------------------------------------------------------------------+
void ClearStatusLabel(void)
{
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
for(int i = 0; i < g_statusLabelLineCount; i++)
{
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
string bg = STATUS_LABEL_PREFIX + (string)i + "BG";
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
string txt = STATUS_LABEL_PREFIX + (string)i + "Text";
fix: correct metric label in training output and improve status panel comments
2026-07-17 21:53:09 -04:00
if(ObjectFind(0, bg) >= 0)
ObjectDelete(0, bg);
fix(status-label): handle multi-line status text with per-line labels OBJ_LABEL does not render embedded '\n' as line breaks, causing multi-line status text to appear as one truncated line. This commit splits the input text by newlines, creates a separate OBJ_LABEL and tightly-fit OBJ_RECTANGLE_LABEL background for each line, ensuring every line remains legible regardless of chart background.
2026-07-17 21:32:55 -04:00
if(ObjectFind(0, txt) >= 0)
ObjectDelete(0, txt);
}
g_statusLabelLineCount = 0;
fix: word-wrap status labels via TextGetSize() to fix clipping and background overflow Replace the static per-character width approximation with dynamic word-wrapping using `TextGetSize()` to measure actual rendered pixel widths. This prevents status label text from being clipped at the chart edge while its oversized background rectangle extended beyond. Each line is now greedily wrapped to fit within the chart pane minus margins, ensuring the measured string matches what is drawn. Also extracts font constant and adds `WrapLineInto()` helper for reuse.
2026-07-17 21:36:44 -04:00
ChartRedraw(0);
refactor(ai): replace Comment() with StatusLabel for training status display Migrate all per-era classification counts, OOS confusion metrics, and training progress output from the legacy Comment() function to a dedicated StatusLabel object. This provides cleaner UI integration and avoids blocking the chart's normal info line during prolonged training cycles. A new include for StatusLabel.mqh has been added, and all related documentation comments updated accordingly.
2026-07-17 21:28:59 -04:00
}
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
//+------------------------------------------------------------------+
refactor(comments): box headers to stdlib length The //| box blocks were excluded from 0b06f8e and 5efdb48 and were what remained: 160 of them ran to 10+ lines, the longest to 88. Compressed to their leading topic sentences - 5 lines for a function header, 8 for a file header - keeping the box format and the standard MQL5 name/author lines verbatim. Verified at the BYTE level this time, across every in-scope file: the list of non-comment lines is byte-identical to HEAD and braces balance. The first check compared a locale-decoded 'git show' against a UTF-8 read and flagged 25 files that had not changed at all - every BOM and every non-ASCII line mismatched. 47,696 -> 40,665 lines in scope; comment share 38% -> 26%. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-22 00:30:14 -04:00
//| ENSEMBLE PANEL (two or more NNs enabled). |
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
//+------------------------------------------------------------------+
#define ENSEMBLE_PANEL_MAX_MEMBERS 6
string g_ensPanelTag[ENSEMBLE_PANEL_MAX_MEMBERS];
string g_ensPanelLine[ENSEMBLE_PANEL_MAX_MEMBERS];
refactor(comments): stdlib comment style across the remaining in-scope files Same pass as 0b06f8e, applied file by file: comment runs of 4+ lines compressed to their leading topic sentences, capped at 4 lines, whole sentences only. Warning sentences (NEVER / MUST / trap / would-have) survive the budget. Every file was checked the same way before committing: the list of non-comment lines is byte-identical to HEAD, and braces balance. No code was touched. Panel/, Enumerations/ and the already-terse System headers needed little or nothing - PooledGate, TradeChecks, BinomialStats and Random came through with no blocks over the threshold at all. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-22 00:25:52 -04:00
//--- ROW OWNERSHIP, so the panel is ordered by the ENSEMBLE's own member order and not by which
//--- member happened to publish first.
fix(indicators+panel): the dead handle is MEASURED now - recreate it; and order the ensemble panel by member, not by who published first THE ANSWER, off the instrumentation added in be39674, first run: ConvLSTM [HYB-2484]: TUNABLE INDICATOR REPORTS NO CALCULATED BARS - 1 tunable indicator(s) enabled and the least-ready answers BarsCalculated()=-1 ... Per-indicator depth: price=33982 MA=-1 ZigZag=33982 ATR=33982 MA=-1 with price, ZigZag and ATR all at full depth. **The handle is INVALID, not short.** Same line on USDJPY (price=50179 MA=-1). Depth was never the problem; the previous session's five theories were all answering the wrong question. And it is per-member, not per-chart: LSTM-2484 ran the 34-candidate auto-tune on that same chart at 15:24:18 and went on to train normally (feature health, 51 features, excursion head) reading the same indicator. Only ConvLSTM's handle - the last member constructed - was dead. WHY is still not established. All four members request ADMovingAverage with identical params, so MT5 hands them the SAME refcounted handle, and the tuner's inner loop is Create-then-IndicatorRelease over exactly that shared handle; that is the obvious suspect and it is NOT yet proven, so this commit does not act on it. 1. IndicatorDepthReport() NOW PRINTS HANDLE NUMBERS, not just depths. "MA=-1" says the handle is dead. "MA=-1(h12)" against another member's "MA=33982 (h12)" says it is the SAME handle and someone released it; "(h-1)" says it was never created. That is the difference between a refcount bug and a creation failure and it is one field. This is the measurement the shared-handle suspicion needs before anyone acts on it. 2. RECREATE A DEAD HANDLE INSTEAD OF SWEEPING AGAINST IT. A member that cannot read its own indicator must rebuild it. RepairDeadIndicatorHandles() re-Creates only the ENABLED tunables reporting BarsCalculated() < 0 - a merely COLD indicator (valid handle, 0 bars) is left alone to warm up the normal way. It does NOT release first: -1 means the terminal no longer knows the handle, so there is nothing to give back, and MT5 recycles handle VALUES so releasing a stale one could decrement whatever now owns that number. 30s cooldown, because every ServableBars() consumer reaches it including live inference on every tick. The feature cache is dropped with it, and the log names before/after depths. Cause-agnostic on purpose. Whatever is killing the handle, sweeping 50,163 bars against a buffer that answers EMPTY_VALUE at every index - then discarding the era and doing it again - is not a recovery. 3. THE SWEEP NOW HOLDS ON A DEAD HANDLE. ServableBars() keeps answering `want` (its contract; live inference and online learning have their own refusal paths and a 0 there reads as "no history at all"). SettledBars() - the training sweep's entry, the one caller that can afford to wait - returns 0 instead, so Train() holds and reports rather than burning a full-history pass it is guaranteed to throw away. A recreated handle is cold, so it primes through the existing settle path on the next call. If the repair fails the member holds indefinitely and says so every minute, and be39674's barrier liveness escape releases the rest of the ensemble after 12 minutes - which is the correct degradation and is exactly what the log shows happening. 4. THE PANEL ROWS WERE ORDERED BY WHO PUBLISHED FIRST. Reported on XAUUSD: LSTM, ConvLSTM, Perceptron, Convolutional instead of Perceptron, Convolutional, LSTM, ConvLSTM. ClaimEnsemblePanelSlot() handed out the next free row on each member's FIRST PublishStatus() call, so the order was a race - the members busy sweeping published before the ones sitting idle at the era barrier, and be39674 sharpened it by (correctly) making a held member stop writing the terse line. Rows are now keyed to m_ensembleIndex, the registration/construction order, which is fixed for the life of the chart. Claimed on every publish rather than once, so it is idempotent and refreshes the tag for a member whose ID was not final when it first published (the config-tag suffix is appended during InitIndicators, after EnsembleMember() registers). Unclaimed rows are skipped by the render and excluded from the model count, so a member that has not published yet leaves no gap and shifts nobody. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-17 15:40:00 -04:00
bool g_ensPanelSlotUsed[ENSEMBLE_PANEL_MAX_MEMBERS];
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
int g_ensPanelUsed = 0;
uint g_ensPanelLastRender = 0;
string g_ensPanelLastText = "";
fix(ensemble): responsive panel + synchronized eras + combined-vote accuracy Four user-reported/requested items, one root cause chain: 1) DEAD CONTROL PANEL in AI_HYBRID mode. All members posted custom event id 1 and handled id 1001, and CExpertCustom broadcasts every chart event to every filter - so each posted event ran a train chunk in ALL N members (N*N chunks per round) and the chart thread never idled long enough to deliver clicks/drags. profiling.csv: 99.45% of time in OnChartEventHandler. Fix: per-instance study-event ids (STUDY_EVENT_ID_BASE + construction order, offset above the Controls library's ON_* codes - id 1 was also ON_DBL_CLICK, so panel double-clicks fired training chunks). ArmStudyEvent() is the single post site; lost-event watchdog replaces the accidental sibling-clears-my-flag rescue. 2) WARM-UP DUPLICATION. The auto-tune sweep is deterministic over identical features/labels, and it ends in the full MI diagnostic suite, which the MI-share gate never intercepted on the sweep path - four members ran four identical ~36s sweep+report blocks. First member publishes outcome (g_ensembleChartTuneDone/Installed/Settings); the rest apply it and skip both. 3) DEINIT STRANDED PANEL+ARROWS (user repro 18:52). Root cause from the log: the 4,500ms budget runs from MetaTrader's stop REQUEST - a heavy autosave in flight ate it, OnDeinit got ~430ms and died in the first member's arrow persist ("Abnormal termination" 432ms in). Fix: early visible-UI sweep (native prefix deletes for status/panel/dialog) right after ClearStatusLabel, and a fast path for still-training models - their arrows are re-rendered every era, so they get one bulk purge instead of scan+atomic-write in the death window. 4) ENSEMBLE FEATURES (user requests): era BARRIER - members advance era by era together; a member ahead of the slowest still-training member declines Train() calls and its chunk budget is donated (TRAIN_TIME_BUDGET_MS = 120/activeTrainers, UI headroom constant). COMBINED-VOTE OOS SCORE - each member's pass-3 scan contributes its adjusted per-bar decision (0.0 on abstain) to a shared row buffer; the last member to finish the era scores the averaged vote vs the mirrored Min_Vote_Open against the same target-before-stop outcomes members grade themselves on, publishing an "Ensemble vote" line on the aggregated panel. Member headlines now carry their lifetime win rate with break-even. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-16 19:06:04 -04:00
//--- The ensemble's COMBINED-vote OOS score line, written once per era by the last member to finish
//--- its pass-3 scan (CExpertSignalAIBase::EnsembleScoreCombinedVote) and rendered under the member
refactor(comments): stdlib comment style across the remaining in-scope files Same pass as 0b06f8e, applied file by file: comment runs of 4+ lines compressed to their leading topic sentences, capped at 4 lines, whole sentences only. Warning sentences (NEVER / MUST / trap / would-have) survive the budget. Every file was checked the same way before committing: the list of non-comment lines is byte-identical to HEAD, and braces balance. No code was touched. Panel/, Enumerations/ and the already-terse System headers needed little or nothing - PooledGate, TradeChecks, BinomialStats and Random came through with no blocks over the threshold at all. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-22 00:25:52 -04:00
//--- lines below.
fix(ensemble): responsive panel + synchronized eras + combined-vote accuracy Four user-reported/requested items, one root cause chain: 1) DEAD CONTROL PANEL in AI_HYBRID mode. All members posted custom event id 1 and handled id 1001, and CExpertCustom broadcasts every chart event to every filter - so each posted event ran a train chunk in ALL N members (N*N chunks per round) and the chart thread never idled long enough to deliver clicks/drags. profiling.csv: 99.45% of time in OnChartEventHandler. Fix: per-instance study-event ids (STUDY_EVENT_ID_BASE + construction order, offset above the Controls library's ON_* codes - id 1 was also ON_DBL_CLICK, so panel double-clicks fired training chunks). ArmStudyEvent() is the single post site; lost-event watchdog replaces the accidental sibling-clears-my-flag rescue. 2) WARM-UP DUPLICATION. The auto-tune sweep is deterministic over identical features/labels, and it ends in the full MI diagnostic suite, which the MI-share gate never intercepted on the sweep path - four members ran four identical ~36s sweep+report blocks. First member publishes outcome (g_ensembleChartTuneDone/Installed/Settings); the rest apply it and skip both. 3) DEINIT STRANDED PANEL+ARROWS (user repro 18:52). Root cause from the log: the 4,500ms budget runs from MetaTrader's stop REQUEST - a heavy autosave in flight ate it, OnDeinit got ~430ms and died in the first member's arrow persist ("Abnormal termination" 432ms in). Fix: early visible-UI sweep (native prefix deletes for status/panel/dialog) right after ClearStatusLabel, and a fast path for still-training models - their arrows are re-rendered every era, so they get one bulk purge instead of scan+atomic-write in the death window. 4) ENSEMBLE FEATURES (user requests): era BARRIER - members advance era by era together; a member ahead of the slowest still-training member declines Train() calls and its chunk budget is donated (TRAIN_TIME_BUDGET_MS = 120/activeTrainers, UI headroom constant). COMBINED-VOTE OOS SCORE - each member's pass-3 scan contributes its adjusted per-bar decision (0.0 on abstain) to a shared row buffer; the last member to finish the era scores the averaged vote vs the mirrored Min_Vote_Open against the same target-before-stop outcomes members grade themselves on, publishing an "Ensemble vote" line on the aggregated panel. Member headlines now carry their lifetime win rate with break-even. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-16 19:06:04 -04:00
string g_ensembleVoteLine = "";
fix(indicators+panel): the dead handle is MEASURED now - recreate it; and order the ensemble panel by member, not by who published first THE ANSWER, off the instrumentation added in be39674, first run: ConvLSTM [HYB-2484]: TUNABLE INDICATOR REPORTS NO CALCULATED BARS - 1 tunable indicator(s) enabled and the least-ready answers BarsCalculated()=-1 ... Per-indicator depth: price=33982 MA=-1 ZigZag=33982 ATR=33982 MA=-1 with price, ZigZag and ATR all at full depth. **The handle is INVALID, not short.** Same line on USDJPY (price=50179 MA=-1). Depth was never the problem; the previous session's five theories were all answering the wrong question. And it is per-member, not per-chart: LSTM-2484 ran the 34-candidate auto-tune on that same chart at 15:24:18 and went on to train normally (feature health, 51 features, excursion head) reading the same indicator. Only ConvLSTM's handle - the last member constructed - was dead. WHY is still not established. All four members request ADMovingAverage with identical params, so MT5 hands them the SAME refcounted handle, and the tuner's inner loop is Create-then-IndicatorRelease over exactly that shared handle; that is the obvious suspect and it is NOT yet proven, so this commit does not act on it. 1. IndicatorDepthReport() NOW PRINTS HANDLE NUMBERS, not just depths. "MA=-1" says the handle is dead. "MA=-1(h12)" against another member's "MA=33982 (h12)" says it is the SAME handle and someone released it; "(h-1)" says it was never created. That is the difference between a refcount bug and a creation failure and it is one field. This is the measurement the shared-handle suspicion needs before anyone acts on it. 2. RECREATE A DEAD HANDLE INSTEAD OF SWEEPING AGAINST IT. A member that cannot read its own indicator must rebuild it. RepairDeadIndicatorHandles() re-Creates only the ENABLED tunables reporting BarsCalculated() < 0 - a merely COLD indicator (valid handle, 0 bars) is left alone to warm up the normal way. It does NOT release first: -1 means the terminal no longer knows the handle, so there is nothing to give back, and MT5 recycles handle VALUES so releasing a stale one could decrement whatever now owns that number. 30s cooldown, because every ServableBars() consumer reaches it including live inference on every tick. The feature cache is dropped with it, and the log names before/after depths. Cause-agnostic on purpose. Whatever is killing the handle, sweeping 50,163 bars against a buffer that answers EMPTY_VALUE at every index - then discarding the era and doing it again - is not a recovery. 3. THE SWEEP NOW HOLDS ON A DEAD HANDLE. ServableBars() keeps answering `want` (its contract; live inference and online learning have their own refusal paths and a 0 there reads as "no history at all"). SettledBars() - the training sweep's entry, the one caller that can afford to wait - returns 0 instead, so Train() holds and reports rather than burning a full-history pass it is guaranteed to throw away. A recreated handle is cold, so it primes through the existing settle path on the next call. If the repair fails the member holds indefinitely and says so every minute, and be39674's barrier liveness escape releases the rest of the ensemble after 12 minutes - which is the correct degradation and is exactly what the log shows happening. 4. THE PANEL ROWS WERE ORDERED BY WHO PUBLISHED FIRST. Reported on XAUUSD: LSTM, ConvLSTM, Perceptron, Convolutional instead of Perceptron, Convolutional, LSTM, ConvLSTM. ClaimEnsemblePanelSlot() handed out the next free row on each member's FIRST PublishStatus() call, so the order was a race - the members busy sweeping published before the ones sitting idle at the era barrier, and be39674 sharpened it by (correctly) making a held member stop writing the terse line. Rows are now keyed to m_ensembleIndex, the registration/construction order, which is fixed for the life of the chart. Claimed on every publish rather than once, so it is idempotent and refreshes the tag for a member whose ID was not final when it first published (the config-tag suffix is appended during InitIndicators, after EnsembleMember() registers). Unclaimed rows are skipped by the render and excluded from the model count, so a member that has not published yet leaves no gap and shifts nobody. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-17 15:40:00 -04:00
//--- index = the caller's ensemble registration index. Idempotent: safe to call on every publish, which
//--- is what keeps the tag current for a member whose ID was not final when it first claimed its row.
int ClaimEnsemblePanelSlot(string tag, int index)
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
{
fix(indicators+panel): the dead handle is MEASURED now - recreate it; and order the ensemble panel by member, not by who published first THE ANSWER, off the instrumentation added in be39674, first run: ConvLSTM [HYB-2484]: TUNABLE INDICATOR REPORTS NO CALCULATED BARS - 1 tunable indicator(s) enabled and the least-ready answers BarsCalculated()=-1 ... Per-indicator depth: price=33982 MA=-1 ZigZag=33982 ATR=33982 MA=-1 with price, ZigZag and ATR all at full depth. **The handle is INVALID, not short.** Same line on USDJPY (price=50179 MA=-1). Depth was never the problem; the previous session's five theories were all answering the wrong question. And it is per-member, not per-chart: LSTM-2484 ran the 34-candidate auto-tune on that same chart at 15:24:18 and went on to train normally (feature health, 51 features, excursion head) reading the same indicator. Only ConvLSTM's handle - the last member constructed - was dead. WHY is still not established. All four members request ADMovingAverage with identical params, so MT5 hands them the SAME refcounted handle, and the tuner's inner loop is Create-then-IndicatorRelease over exactly that shared handle; that is the obvious suspect and it is NOT yet proven, so this commit does not act on it. 1. IndicatorDepthReport() NOW PRINTS HANDLE NUMBERS, not just depths. "MA=-1" says the handle is dead. "MA=-1(h12)" against another member's "MA=33982 (h12)" says it is the SAME handle and someone released it; "(h-1)" says it was never created. That is the difference between a refcount bug and a creation failure and it is one field. This is the measurement the shared-handle suspicion needs before anyone acts on it. 2. RECREATE A DEAD HANDLE INSTEAD OF SWEEPING AGAINST IT. A member that cannot read its own indicator must rebuild it. RepairDeadIndicatorHandles() re-Creates only the ENABLED tunables reporting BarsCalculated() < 0 - a merely COLD indicator (valid handle, 0 bars) is left alone to warm up the normal way. It does NOT release first: -1 means the terminal no longer knows the handle, so there is nothing to give back, and MT5 recycles handle VALUES so releasing a stale one could decrement whatever now owns that number. 30s cooldown, because every ServableBars() consumer reaches it including live inference on every tick. The feature cache is dropped with it, and the log names before/after depths. Cause-agnostic on purpose. Whatever is killing the handle, sweeping 50,163 bars against a buffer that answers EMPTY_VALUE at every index - then discarding the era and doing it again - is not a recovery. 3. THE SWEEP NOW HOLDS ON A DEAD HANDLE. ServableBars() keeps answering `want` (its contract; live inference and online learning have their own refusal paths and a 0 there reads as "no history at all"). SettledBars() - the training sweep's entry, the one caller that can afford to wait - returns 0 instead, so Train() holds and reports rather than burning a full-history pass it is guaranteed to throw away. A recreated handle is cold, so it primes through the existing settle path on the next call. If the repair fails the member holds indefinitely and says so every minute, and be39674's barrier liveness escape releases the rest of the ensemble after 12 minutes - which is the correct degradation and is exactly what the log shows happening. 4. THE PANEL ROWS WERE ORDERED BY WHO PUBLISHED FIRST. Reported on XAUUSD: LSTM, ConvLSTM, Perceptron, Convolutional instead of Perceptron, Convolutional, LSTM, ConvLSTM. ClaimEnsemblePanelSlot() handed out the next free row on each member's FIRST PublishStatus() call, so the order was a race - the members busy sweeping published before the ones sitting idle at the era barrier, and be39674 sharpened it by (correctly) making a held member stop writing the terse line. Rows are now keyed to m_ensembleIndex, the registration/construction order, which is fixed for the life of the chart. Claimed on every publish rather than once, so it is idempotent and refreshes the tag for a member whose ID was not final when it first published (the config-tag suffix is appended during InitIndicators, after EnsembleMember() registers). Unclaimed rows are skipped by the render and excluded from the model count, so a member that has not published yet leaves no gap and shifts nobody. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-17 15:40:00 -04:00
int slot = index;
//--- Unregistered or out of range (should not happen for a real member): fall back to the first free
//--- row rather than dropping the model off the panel entirely.
if(slot < 0 || slot >= ENSEMBLE_PANEL_MAX_MEMBERS)
{
slot = -1;
for(int i = 0; i < ENSEMBLE_PANEL_MAX_MEMBERS; i++)
if(!g_ensPanelSlotUsed[i])
{
slot = i;
break;
}
if(slot < 0)
slot = ENSEMBLE_PANEL_MAX_MEMBERS - 1;
}
if(!g_ensPanelSlotUsed[slot])
{
g_ensPanelSlotUsed[slot] = true;
g_ensPanelLine[slot] = "";
}
if(tag != "")
g_ensPanelTag[slot] = tag;
//--- High-water mark of CLAIMED rows, not a running count - rows are addressed by index now, so a
//--- member that has not published yet leaves a gap the render skips rather than shifting everyone up.
if(slot + 1 > g_ensPanelUsed)
g_ensPanelUsed = slot + 1;
return slot;
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
}
void PublishEnsembleStatus(int slot, string headline, bool force = false)
{
fix(indicators+panel): the dead handle is MEASURED now - recreate it; and order the ensemble panel by member, not by who published first THE ANSWER, off the instrumentation added in be39674, first run: ConvLSTM [HYB-2484]: TUNABLE INDICATOR REPORTS NO CALCULATED BARS - 1 tunable indicator(s) enabled and the least-ready answers BarsCalculated()=-1 ... Per-indicator depth: price=33982 MA=-1 ZigZag=33982 ATR=33982 MA=-1 with price, ZigZag and ATR all at full depth. **The handle is INVALID, not short.** Same line on USDJPY (price=50179 MA=-1). Depth was never the problem; the previous session's five theories were all answering the wrong question. And it is per-member, not per-chart: LSTM-2484 ran the 34-candidate auto-tune on that same chart at 15:24:18 and went on to train normally (feature health, 51 features, excursion head) reading the same indicator. Only ConvLSTM's handle - the last member constructed - was dead. WHY is still not established. All four members request ADMovingAverage with identical params, so MT5 hands them the SAME refcounted handle, and the tuner's inner loop is Create-then-IndicatorRelease over exactly that shared handle; that is the obvious suspect and it is NOT yet proven, so this commit does not act on it. 1. IndicatorDepthReport() NOW PRINTS HANDLE NUMBERS, not just depths. "MA=-1" says the handle is dead. "MA=-1(h12)" against another member's "MA=33982 (h12)" says it is the SAME handle and someone released it; "(h-1)" says it was never created. That is the difference between a refcount bug and a creation failure and it is one field. This is the measurement the shared-handle suspicion needs before anyone acts on it. 2. RECREATE A DEAD HANDLE INSTEAD OF SWEEPING AGAINST IT. A member that cannot read its own indicator must rebuild it. RepairDeadIndicatorHandles() re-Creates only the ENABLED tunables reporting BarsCalculated() < 0 - a merely COLD indicator (valid handle, 0 bars) is left alone to warm up the normal way. It does NOT release first: -1 means the terminal no longer knows the handle, so there is nothing to give back, and MT5 recycles handle VALUES so releasing a stale one could decrement whatever now owns that number. 30s cooldown, because every ServableBars() consumer reaches it including live inference on every tick. The feature cache is dropped with it, and the log names before/after depths. Cause-agnostic on purpose. Whatever is killing the handle, sweeping 50,163 bars against a buffer that answers EMPTY_VALUE at every index - then discarding the era and doing it again - is not a recovery. 3. THE SWEEP NOW HOLDS ON A DEAD HANDLE. ServableBars() keeps answering `want` (its contract; live inference and online learning have their own refusal paths and a 0 there reads as "no history at all"). SettledBars() - the training sweep's entry, the one caller that can afford to wait - returns 0 instead, so Train() holds and reports rather than burning a full-history pass it is guaranteed to throw away. A recreated handle is cold, so it primes through the existing settle path on the next call. If the repair fails the member holds indefinitely and says so every minute, and be39674's barrier liveness escape releases the rest of the ensemble after 12 minutes - which is the correct degradation and is exactly what the log shows happening. 4. THE PANEL ROWS WERE ORDERED BY WHO PUBLISHED FIRST. Reported on XAUUSD: LSTM, ConvLSTM, Perceptron, Convolutional instead of Perceptron, Convolutional, LSTM, ConvLSTM. ClaimEnsemblePanelSlot() handed out the next free row on each member's FIRST PublishStatus() call, so the order was a race - the members busy sweeping published before the ones sitting idle at the era barrier, and be39674 sharpened it by (correctly) making a held member stop writing the terse line. Rows are now keyed to m_ensembleIndex, the registration/construction order, which is fixed for the life of the chart. Claimed on every publish rather than once, so it is idempotent and refreshes the tag for a member whose ID was not final when it first published (the config-tag suffix is appended during InitIndicators, after EnsembleMember() registers). Unclaimed rows are skipped by the render and excluded from the model count, so a member that has not published yet leaves no gap and shifts nobody. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-17 15:40:00 -04:00
if(slot < 0 || slot >= g_ensPanelUsed || !g_ensPanelSlotUsed[slot])
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
return;
g_ensPanelLine[slot] = headline;
fix(indicators+panel): the dead handle is MEASURED now - recreate it; and order the ensemble panel by member, not by who published first THE ANSWER, off the instrumentation added in be39674, first run: ConvLSTM [HYB-2484]: TUNABLE INDICATOR REPORTS NO CALCULATED BARS - 1 tunable indicator(s) enabled and the least-ready answers BarsCalculated()=-1 ... Per-indicator depth: price=33982 MA=-1 ZigZag=33982 ATR=33982 MA=-1 with price, ZigZag and ATR all at full depth. **The handle is INVALID, not short.** Same line on USDJPY (price=50179 MA=-1). Depth was never the problem; the previous session's five theories were all answering the wrong question. And it is per-member, not per-chart: LSTM-2484 ran the 34-candidate auto-tune on that same chart at 15:24:18 and went on to train normally (feature health, 51 features, excursion head) reading the same indicator. Only ConvLSTM's handle - the last member constructed - was dead. WHY is still not established. All four members request ADMovingAverage with identical params, so MT5 hands them the SAME refcounted handle, and the tuner's inner loop is Create-then-IndicatorRelease over exactly that shared handle; that is the obvious suspect and it is NOT yet proven, so this commit does not act on it. 1. IndicatorDepthReport() NOW PRINTS HANDLE NUMBERS, not just depths. "MA=-1" says the handle is dead. "MA=-1(h12)" against another member's "MA=33982 (h12)" says it is the SAME handle and someone released it; "(h-1)" says it was never created. That is the difference between a refcount bug and a creation failure and it is one field. This is the measurement the shared-handle suspicion needs before anyone acts on it. 2. RECREATE A DEAD HANDLE INSTEAD OF SWEEPING AGAINST IT. A member that cannot read its own indicator must rebuild it. RepairDeadIndicatorHandles() re-Creates only the ENABLED tunables reporting BarsCalculated() < 0 - a merely COLD indicator (valid handle, 0 bars) is left alone to warm up the normal way. It does NOT release first: -1 means the terminal no longer knows the handle, so there is nothing to give back, and MT5 recycles handle VALUES so releasing a stale one could decrement whatever now owns that number. 30s cooldown, because every ServableBars() consumer reaches it including live inference on every tick. The feature cache is dropped with it, and the log names before/after depths. Cause-agnostic on purpose. Whatever is killing the handle, sweeping 50,163 bars against a buffer that answers EMPTY_VALUE at every index - then discarding the era and doing it again - is not a recovery. 3. THE SWEEP NOW HOLDS ON A DEAD HANDLE. ServableBars() keeps answering `want` (its contract; live inference and online learning have their own refusal paths and a 0 there reads as "no history at all"). SettledBars() - the training sweep's entry, the one caller that can afford to wait - returns 0 instead, so Train() holds and reports rather than burning a full-history pass it is guaranteed to throw away. A recreated handle is cold, so it primes through the existing settle path on the next call. If the repair fails the member holds indefinitely and says so every minute, and be39674's barrier liveness escape releases the rest of the ensemble after 12 minutes - which is the correct degradation and is exactly what the log shows happening. 4. THE PANEL ROWS WERE ORDERED BY WHO PUBLISHED FIRST. Reported on XAUUSD: LSTM, ConvLSTM, Perceptron, Convolutional instead of Perceptron, Convolutional, LSTM, ConvLSTM. ClaimEnsemblePanelSlot() handed out the next free row on each member's FIRST PublishStatus() call, so the order was a race - the members busy sweeping published before the ones sitting idle at the era barrier, and be39674 sharpened it by (correctly) making a held member stop writing the terse line. Rows are now keyed to m_ensembleIndex, the registration/construction order, which is fixed for the life of the chart. Claimed on every publish rather than once, so it is idempotent and refreshes the tag for a member whose ID was not final when it first published (the config-tag suffix is appended during InitIndicators, after EnsembleMember() registers). Unclaimed rows are skipped by the render and excluded from the model count, so a member that has not published yet leaves no gap and shifts nobody. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-17 15:40:00 -04:00
int shown = 0;
for(int i = 0; i < g_ensPanelUsed; i++)
if(g_ensPanelSlotUsed[i])
shown++;
string s = "AI ensemble - " + (string)shown + " independent models, one line each";
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
for(int i = 0; i < g_ensPanelUsed; i++)
fix(indicators+panel): the dead handle is MEASURED now - recreate it; and order the ensemble panel by member, not by who published first THE ANSWER, off the instrumentation added in be39674, first run: ConvLSTM [HYB-2484]: TUNABLE INDICATOR REPORTS NO CALCULATED BARS - 1 tunable indicator(s) enabled and the least-ready answers BarsCalculated()=-1 ... Per-indicator depth: price=33982 MA=-1 ZigZag=33982 ATR=33982 MA=-1 with price, ZigZag and ATR all at full depth. **The handle is INVALID, not short.** Same line on USDJPY (price=50179 MA=-1). Depth was never the problem; the previous session's five theories were all answering the wrong question. And it is per-member, not per-chart: LSTM-2484 ran the 34-candidate auto-tune on that same chart at 15:24:18 and went on to train normally (feature health, 51 features, excursion head) reading the same indicator. Only ConvLSTM's handle - the last member constructed - was dead. WHY is still not established. All four members request ADMovingAverage with identical params, so MT5 hands them the SAME refcounted handle, and the tuner's inner loop is Create-then-IndicatorRelease over exactly that shared handle; that is the obvious suspect and it is NOT yet proven, so this commit does not act on it. 1. IndicatorDepthReport() NOW PRINTS HANDLE NUMBERS, not just depths. "MA=-1" says the handle is dead. "MA=-1(h12)" against another member's "MA=33982 (h12)" says it is the SAME handle and someone released it; "(h-1)" says it was never created. That is the difference between a refcount bug and a creation failure and it is one field. This is the measurement the shared-handle suspicion needs before anyone acts on it. 2. RECREATE A DEAD HANDLE INSTEAD OF SWEEPING AGAINST IT. A member that cannot read its own indicator must rebuild it. RepairDeadIndicatorHandles() re-Creates only the ENABLED tunables reporting BarsCalculated() < 0 - a merely COLD indicator (valid handle, 0 bars) is left alone to warm up the normal way. It does NOT release first: -1 means the terminal no longer knows the handle, so there is nothing to give back, and MT5 recycles handle VALUES so releasing a stale one could decrement whatever now owns that number. 30s cooldown, because every ServableBars() consumer reaches it including live inference on every tick. The feature cache is dropped with it, and the log names before/after depths. Cause-agnostic on purpose. Whatever is killing the handle, sweeping 50,163 bars against a buffer that answers EMPTY_VALUE at every index - then discarding the era and doing it again - is not a recovery. 3. THE SWEEP NOW HOLDS ON A DEAD HANDLE. ServableBars() keeps answering `want` (its contract; live inference and online learning have their own refusal paths and a 0 there reads as "no history at all"). SettledBars() - the training sweep's entry, the one caller that can afford to wait - returns 0 instead, so Train() holds and reports rather than burning a full-history pass it is guaranteed to throw away. A recreated handle is cold, so it primes through the existing settle path on the next call. If the repair fails the member holds indefinitely and says so every minute, and be39674's barrier liveness escape releases the rest of the ensemble after 12 minutes - which is the correct degradation and is exactly what the log shows happening. 4. THE PANEL ROWS WERE ORDERED BY WHO PUBLISHED FIRST. Reported on XAUUSD: LSTM, ConvLSTM, Perceptron, Convolutional instead of Perceptron, Convolutional, LSTM, ConvLSTM. ClaimEnsemblePanelSlot() handed out the next free row on each member's FIRST PublishStatus() call, so the order was a race - the members busy sweeping published before the ones sitting idle at the era barrier, and be39674 sharpened it by (correctly) making a held member stop writing the terse line. Rows are now keyed to m_ensembleIndex, the registration/construction order, which is fixed for the life of the chart. Claimed on every publish rather than once, so it is idempotent and refreshes the tag for a member whose ID was not final when it first published (the config-tag suffix is appended during InitIndicators, after EnsembleMember() registers). Unclaimed rows are skipped by the render and excluded from the model count, so a member that has not published yet leaves no gap and shifts nobody. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-17 15:40:00 -04:00
if(g_ensPanelSlotUsed[i])
s += "\n" + g_ensPanelTag[i] + ": " + (g_ensPanelLine[i] == "" ? "starting..." : g_ensPanelLine[i]);
fix(ensemble): responsive panel + synchronized eras + combined-vote accuracy Four user-reported/requested items, one root cause chain: 1) DEAD CONTROL PANEL in AI_HYBRID mode. All members posted custom event id 1 and handled id 1001, and CExpertCustom broadcasts every chart event to every filter - so each posted event ran a train chunk in ALL N members (N*N chunks per round) and the chart thread never idled long enough to deliver clicks/drags. profiling.csv: 99.45% of time in OnChartEventHandler. Fix: per-instance study-event ids (STUDY_EVENT_ID_BASE + construction order, offset above the Controls library's ON_* codes - id 1 was also ON_DBL_CLICK, so panel double-clicks fired training chunks). ArmStudyEvent() is the single post site; lost-event watchdog replaces the accidental sibling-clears-my-flag rescue. 2) WARM-UP DUPLICATION. The auto-tune sweep is deterministic over identical features/labels, and it ends in the full MI diagnostic suite, which the MI-share gate never intercepted on the sweep path - four members ran four identical ~36s sweep+report blocks. First member publishes outcome (g_ensembleChartTuneDone/Installed/Settings); the rest apply it and skip both. 3) DEINIT STRANDED PANEL+ARROWS (user repro 18:52). Root cause from the log: the 4,500ms budget runs from MetaTrader's stop REQUEST - a heavy autosave in flight ate it, OnDeinit got ~430ms and died in the first member's arrow persist ("Abnormal termination" 432ms in). Fix: early visible-UI sweep (native prefix deletes for status/panel/dialog) right after ClearStatusLabel, and a fast path for still-training models - their arrows are re-rendered every era, so they get one bulk purge instead of scan+atomic-write in the death window. 4) ENSEMBLE FEATURES (user requests): era BARRIER - members advance era by era together; a member ahead of the slowest still-training member declines Train() calls and its chunk budget is donated (TRAIN_TIME_BUDGET_MS = 120/activeTrainers, UI headroom constant). COMBINED-VOTE OOS SCORE - each member's pass-3 scan contributes its adjusted per-bar decision (0.0 on abstain) to a shared row buffer; the last member to finish the era scores the averaged vote vs the mirrored Min_Vote_Open against the same target-before-stop outcomes members grade themselves on, publishing an "Ensemble vote" line on the aggregated panel. Member headlines now carry their lifetime win rate with break-even. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-16 19:06:04 -04:00
if(g_ensembleVoteLine != "")
s += "\n" + g_ensembleVoteLine;
feat(ui): one aggregated status panel for the AI_HYBRID ensemble All four ensemble members previously wrote their full multi-line panels to the SAME global label objects - an ensemble chart would flicker between four stacked panels covering the chart side (user request: aggregate). Every AI-side SetStatusLabel call site now routes through CExpertSignalAIBase::PublishStatus - solo charts draw the full panel exactly as before; an ensemble member claims a slot and contributes only its HEADLINE to one combined block ("HYBRID ensemble - N models", then one line per model; the live line leads with the model current signal). The combined render skips unchanged text and enforces its own minimum redraw interval so four publishers cannot multiply ChartRedraw() cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-15 16:54:43 -04:00
//--- unchanged text needs no relayout; changed text still respects a minimum redraw interval
if(s == g_ensPanelLastText)
return;
uint now = GetTickCount();
if(!force && g_ensPanelLastRender != 0 && now - g_ensPanelLastRender < 300)
return;
g_ensPanelLastText = s;
g_ensPanelLastRender = now;
SetStatusLabel(s);
}
Copy permalink