Process bidirectional text in conformance to Unicode standards

CHANGELOG.md

@@ -19,6 +19,7 @@ This can also be enabled programmatically with `warnings.simplefilter('default',
 ## [2.7.8] - Not released yet
 ### Added
 * support for `<path>` elements in SVG `<clipPath>` elements
+* support for `bidirectional` text shaping - thanks to @andersonhc
 * documentation on how to combine `fpdf2` with [mistletoe](https://pypi.org/project/kaleido/) in order to [generate PDF documents from Markdown (link)](https://py-pdf.github.io/fpdf2/CombineWithMistletoeoToUseMarkdown.html)
 * tutorial in Dutch: [Handleiding](https://py-pdf.github.io/fpdf2/Tutorial-nl.md) - thanks to @Polderrider
 * support for `Table` cells that span multiple rows via the `rowspan` attribute, which can be combined with `colspan` - thanks to @mjasperse

docs/Unicode.md

@@ -84,17 +84,9 @@ Many non-latin writing systems have complex ways to combine characters, ligature
 To make sure those scripts to be rendered correctly, [text shaping](TextShaping.html) must be enabled with `.set_text_shaping(True)`. 
 
 
-### Right-to-Left & Arabic Script workaround ###
+### Right-to-Left scripts ###
 
-Arabic, Hebrew and other scripts written right-to-left (RTL) should work correctly when text is added that only contains one script at a time. As of release 2.7.6, more complete support for mixing RTL and LTR text is being worked on.
-In the mean time, there is a temporary solution for Arabic and other RTL scripts using two additional libraries `python-bidi` and `arabic-reshaper`. It works for most languages; only a few (rare) Arabic characters aren't supported. Using it on other scripts (eg. when the input is unknown or mixed scripts) does not affect them:
-```python
-from arabic_reshaper import reshape
-from bidi.algorithm import get_display
-
-some_text = 'اَلْعَرَبِيَّةُכַּף סוֹפִית'
-fixed_text = get_display(reshape(some_text))
-```
+When [text shaping](TextShaping.html) is enabled, `fpdf2` will apply the [Unicode Bidirectional Algorithm](https://www.unicode.org/reports/tr9/) to render correctly any text, including bidirectional (mix of right-to-left and left-to-right scripts).
 
 ## Example ##
 

fpdf/fonts.py

@@ -286,17 +286,16 @@ def perform_harfbuzz_shaping(self, text, font_size_pt, text_shaping_parms):
         buf = hb.Buffer()
         buf.cluster_level = 1
         buf.add_str("".join(text))
+        buf.guess_segment_properties()
         features = text_shaping_parms["features"]
-        if (
-            text_shaping_parms["direction"]
-            or text_shaping_parms["script"]
-            or text_shaping_parms["language"]
-        ):
-            buf.direction = text_shaping_parms["direction"]
+        if text_shaping_parms["fragment_direction"]:
+            buf.direction = (
+                "RTL" if text_shaping_parms["fragment_direction"] == "R" else "LTR"
+            )
+        if text_shaping_parms["script"]:
             buf.script = text_shaping_parms["script"]
+        if text_shaping_parms["language"]:
             buf.language = text_shaping_parms["language"]
-        else:
-            buf.guess_segment_properties()
         hb.shape(self.hbfont, buf, features)
         return buf.glyph_infos, buf.glyph_positions