The saving code has been refactored into one class per format with common
methods for each part of the saving process. This should make it easier to
add further formats, and inheriting a format class can allow implementing a
different format that is largely based on an existing format.
The loading code is also in one class per format but is largely the same as
what it was before.
Both classes have the ability to warn the user if any issues or errors are
or will be encountered when loading or saving.
TTI files are now written with CR/LF line endings on all platforms as a
result of using binary file writing for all formats, previously Linux builds
would save TTI files with just LF line endings. TTI files are still loaded
with readLine() and trimmed() which can cope with either type of line ending.
Experimental support for loading and exporting EP1 and HMS SD-Teletext htt
formats has been added. The htt format inherits from the t42 format as the
format appears to be largely the same except for the bits being reversed
within each byte and the clock run-in and framing code added before each
packet.
Actually this is a workaround: it does not attempt to copy construct
unhandled packets, to avoid a bug in the base class where pointers were
copy constructed without copying the contents.
A proper fix will need refactoring of the packet storage code.
These modes can be used to see the inner workings of colourful artworks that
frequently use "foreground colour, new background, foreground colour",
particularly to show exactly which colours are represented by set or clear
sixels.
One mode renders all characters in white on a black background and the other
mode renders black characters on a white background. The latter could be used
to save ink if the resulting image is printed.
Flash is suppressed in these mono modes, this may or may not change.
Qt 5 has had to be dropped as CMakeLists.txt uses qt_* specific commands
which are implemented in Qt 6 onwards.
cmake --install is only working on Linux at the moment; it installs the
executable and the bundled example tti files into /usr/local/share/doc
The teletext decoder with its document storage has been moved to a
subdirectory with the intention of making it possible to use in other
projects, but some further separation of editing stuff will be needed.
When a selection box is dragged, holding Ctrl while pressing the arrow
keys will shift the mosaics one sixel at a time.
If a shifting sixel encounters a non-mosaic cell it will pass through to
the other side of it.
For cut and copy sixels are converted to a monochrome image; unset sixels
are black and set sixels are white. Alphanumerics and control codes are
ignored. Held mosaics are not considered.
For paste a clipboard with only an image is converted to black and white
for unset and set sixels. This only works reliably with fully-black and
fully-white pixels.
This reverts commit 7cd632b4e2.
The commit would cause the font icons to be inverted sometimes. This
reversion is altered to return the font bitmaps as a QPixmap instead of a
QBitmap to be compatible with Qt6.
Amendment to 7b59c78. This should prevent spurious NOS character sets from
appearing in the widgets, and should also prevent the G0 and G2 character
sets from getting mixed up in the widgets when those triplets are within
object definitions.
The G0 and G2 character set in the widgets may still be incorrect in corner
cases of objects invoked within areas of the page where the character set
has been changed either with the ESC/Switch spacing attribute or with the
"modified G0/G2 character set designation" triplet.
A complete rewrite of decode.cpp and decode.h to make it easier to follow
and maintain. Rather than a set of "layers" descended from C++ classes, the
for-loop across the rows and columns explicitly lays out the order of
processing the Level 1 characters and attributes, the X/26 enhancements
that affect the page directly as well as those within each Invocation of an
Object, followed by selecting which Invoked Object to place in the character
cell or in the absence of an Object the underlying page fragment will be
placed in the cell.
The new decoder has the following improvements over the old...
- Local Enhancement Data has priority over Active Objects.
- Active Objects can set Full Screen and Full Row colours.
- Incremental/decremental flash phases are tracked correctly within Objects.
- X/26 characters overwriting the bottom half of Level 1 Double Height rows.
- Interaction between the underlying page and Objects where characters of
different sizes overlap.
This is needed for the upcoming decoder write as decode.cpp will track the
Active Position from the internal data of x26triplets.
If decode.cpp is rewritten again to track the Active Position independently
of x26triplets this commit may be reverted, but then again tracking how the
Active Position moves on both levels 1.5 and 2.5+ may be useful in terms of
warning the user of differing results on different decoder levels.
Previously every character on the page was drawn up to six times. Now the
page is drawn once, quickly copied up to five times and then only the
flashing characters are redrawn on the copies.
Gavin MacGregor