fix(steami_screen): Improve text scaling using pixel-by-pixel framebuf.

[MatteoCnda1]

Summary

Improve _draw_scaled_text() to produce a true pixel-scale zoom instead of a bold offset effect. Closes #369

Changes

• Implemented approach 3 (backend-specific): added draw_scaled_text() directly in SSD1327Display (steami_screen/ssd1327.py)
• The method renders each character into a temporary 8x8 MONO_HLSB framebuf, reads each lit pixel with framebuf.pixel(), and draws a scale x scale filled rectangle for each lit pixel
• _draw_scaled_text() in device.py already checks hasattr(self._d, 'draw_scaled_text') and delegates to the backend — no changes needed in device.py
• This approach allows each backend (SSD1327, GC9A01, etc.) to implement its own optimized scaling strategy

Checklist

• ruff check passes
• python -m pytest tests/ -k mock -v passes (no mock test broken)
• Tested on hardware (STM32WB55 / STeaMi board)
• README updated (if adding/changing public API) — N/A, internal method only
• Examples added/updated — N/A, existing examples benefit automatically
• Commit messages follow <scope>: <Description.> format

[Copilot]

Pull request overview

Improves scaled text rendering for the SSD1327 backend in steami_screen by implementing backend-specific pixel-accurate scaling, and updates documentation to reflect the improved behavior (closes #369).

Changes:

• Added SSD1327Display.draw_scaled_text() that scales glyphs by rendering to a temporary mono framebuffer and expanding lit pixels into scale x scale blocks.
• Updated SSD1327 wrapper internals (quotes/hasattr calls) to support the new scaling path.
• Updated lib/steami_screen/README.md note to describe true pixel-scale zoom behavior on SSD1327.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
lib/steami_screen/steami_screen/ssd1327.py	Adds backend-specific scaled-text rendering via per-pixel expansion from a temporary mono framebuffer.
lib/steami_screen/README.md	Updates user-facing note about text scaling behavior on SSD1327.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Importing framebuf at module import time can make steami_screen fail to import in CPython environments where framebuf isn’t available (unless callers pre-inject a stub). Consider moving this import inside draw_scaled_text() (or guarding it with a try/except) so only the scaled-text path requires framebuf.

[Copilot]

draw_scaled_text() allocates a new bytearray and FrameBuffer on every call. On MicroPython this can add GC pressure and jitter during UI updates. Consider caching the buffer/framebuf on self (lazy-init once) and reusing it across calls.

[Copilot]

The inner pixel loops repeatedly call hasattr(self._raw, "fill_rect") for every lit pixel. Hoist this check once (e.g., pick a local fill_rect function) before the loops to reduce overhead in tight rendering code.

[Copilot]

This implementation can issue up to 64 fill_rect() calls per character (one per set pixel), which is typically slow on MicroPython displays. A common optimization is to coalesce consecutive lit pixels into longer horizontal runs (one fill_rect per run) or build a scaled glyph buffer and blit it, to drastically cut per-character draw calls.

Suggested change

	reads each lit pixel, and draws a scale x scale filled rectangle.
	This produces a true pixel-scale zoom instead of a bold offset effect.
	"""
	gray = rgb_to_gray4(color)
	char_buf = bytearray(8)
	fb = framebuf.FrameBuffer(char_buf, 8, 8, framebuf.MONO_HLSB)
	cx = x
	for char in text:
	fb.fill(0)
	fb.text(char, 0, 0, 1)
	for py in range(8):
	for px in range(8):
	if fb.pixel(px, py):
	if hasattr(self._raw, "fill_rect"):
	self._raw.fill_rect(
	cx + px * scale,
	y + py * scale,
	scale,
	scale,
	gray,
	)
	else:
	self._raw.framebuf.fill_rect(
	cx + px * scale,
	y + py * scale,
	scale,
	scale,
	gray,
	)
	reads each lit pixel, and draws scaled filled rectangles per
	horizontal run of lit pixels. This produces a true pixel-scale zoom
	instead of a bold offset effect while reducing display draw calls.
	"""
	gray = rgb_to_gray4(color)
	char_buf = bytearray(8)
	fb = framebuf.FrameBuffer(char_buf, 8, 8, framebuf.MONO_HLSB)
	if hasattr(self._raw, "fill_rect"):
	fill_rect = self._raw.fill_rect
	else:
	fill_rect = self._raw.framebuf.fill_rect
	cx = x
	for char in text:
	fb.fill(0)
	fb.text(char, 0, 0, 1)
	for py in range(8):
	px = 0
	while px < 8:
	while px < 8 and not fb.pixel(px, py):
	px += 1
	if px >= 8:
	break
	run_start = px
	while px < 8 and fb.pixel(px, py):
	px += 1
	fill_rect(
	cx + run_start * scale,
	y + py * scale,
	(px - run_start) * scale,
	scale,
	gray,
	)

[Copilot]

The note refers specifically to scale=2, but the SSD1327 backend implementation supports arbitrary scale > 1. Consider rewording to scale>1 to avoid implying other scale factors won’t render as true pixel scaling.

Suggested change

	Note: `scale=2` produces a true pixel-scale zoom on SSD1327 displays via pixel-by-pixel
	Note: `scale>1` produces a true pixel-scale zoom on SSD1327 displays via pixel-by-pixel

[nedseb]

Bonne implémentation Matteo. L'approche framebuf 8x8 par caractère est la bonne — c'est la façon standard de faire du scaling sans accès direct au bitmap de la font MicroPython.

J'ai poussé un commit de review (716acdb) qui adresse les 5 commentaires Copilot :

1. import framebuf déféré — déplacé du top-level dans draw_scaled_text() pour que steami_screen reste importable en CPython (tests, stubs IDE, screenshots Pillow).

2. Dispatch fill_rect résolu une seule fois — hasattr(self._raw, 'fill_rect') était appelé à chaque pixel allumé (jusqu'à 64× par caractère × nombre de caractères). Maintenant résolu une fois au __init__ via self._fill_rect_raw, et caché en local blit dans la boucle.

3. Simplification de fill_rect() — la méthode publique utilise directement self._fill_rect_raw, supprimant la duplication du dispatch.

4. Note README — corrigée : le scaling fonctionne pour tout scale > 1, pas seulement scale=2. Mention du fallback bold offset pour les backends sans draw_scaled_text().

Branche rebasée sur main. Prêt à merger.

[semantic-release-updater]

🎉 This PR is included in version 0.20.2 🎉

The release is available on:

• v0.20.2
• GitHub release

Your semantic-release bot 📦🚀