Merge branch 'pr/1712'

CHANGES

@@ -26,6 +26,8 @@ Pint Changelog
 - Parse '°' along with previous text, rather than adding a space,
   allowing, eg 'Δ°C' as a unit.
   (PR #1729)
+- Improved escaping of special characters for LaTeX format
+  (PR #1712)
 - Avoid addition of spurious trailing zeros when converting units and non-int-type is
   Decimal (PR #1625).
 - Implementation for numpy.delete added for Quantity.
@@ -34,6 +36,7 @@ Pint Changelog
   (PR #1722)
 - Added Townsend unit
   (PR #1738)
+
 
 ### Breaking Changes
 

pint/formatting.py

@@ -10,6 +10,7 @@
 
 from __future__ import annotations
 
+import functools
 import re
 import warnings
 from typing import Callable, Dict
@@ -178,10 +179,26 @@ def format_pretty(unit, registry, **options):
     )
 
 
+def latex_escape(string):
+    """
+    Prepend characters that have a special meaning in LaTeX with a backslash.
+    """
+    return functools.reduce(
+        lambda s, m: re.sub(m[0], m[1], s),
+        (
+            (r"[\\]", r"\\textbackslash "),
+            (r"[~]", r"\\textasciitilde "),
+            (r"[\^]", r"\\textasciicircum "),
+            (r"([&%$#_{}])", r"\\\1"),
+        ),
+        str(string),
+    )
+
+
 @register_unit_format("L")
 def format_latex(unit, registry, **options):
     preprocessed = {
-        r"\mathrm{{{}}}".format(u.replace("_", r"\_")): p for u, p in unit.items()
+        r"\mathrm{{{}}}".format(latex_escape(u)): p for u, p in unit.items()
     }
     formatted = formatter(
         preprocessed.items(),

pint/testsuite/test_unit.py

@@ -54,6 +54,33 @@ def test_unit_formatting(self, subtests):
             with subtests.test(spec):
                 assert spec.format(x) == result
 
+    def test_latex_escaping(self, subtests):
+        ureg = UnitRegistry()
+        ureg.define(r"percent = 1e-2 = %")
+        x = ureg.Unit(UnitsContainer(percent=1))
+        for spec, result in {
+            "L": r"\mathrm{percent}",
+            "L~": r"\mathrm{\%}",
+            "Lx": r"\si[]{\percent}",
+            "Lx~": r"\si[]{\%}",
+        }.items():
+            with subtests.test(spec):
+                ureg.default_format = spec
+                assert f"{x}" == result, f"Failed for {spec}, {result}"
+        # no '#' here as it's a comment char when define()ing new units
+        ureg.define(r"weirdunit = 1 = \~_^&%$_{}")
+        x = ureg.Unit(UnitsContainer(weirdunit=1))
+        for spec, result in {
+            "L": r"\mathrm{weirdunit}",
+            "L~": r"\mathrm{\textbackslash \textasciitilde \_\textasciicircum \&\%\$\_\{\}}",
+            "Lx": r"\si[]{\weirdunit}",
+            # TODO: Currently makes \si[]{\\~_^&%$_{}} (broken). What do we even want this to be?
+            # "Lx~": r"\si[]{\textbackslash \textasciitilde \_\textasciicircum \&\%\$\_\{\}}",
+        }.items():
+            with subtests.test(spec):
+                ureg.default_format = spec
+                assert f"{x}" == result, f"Failed for {spec}, {result}"
+
     def test_unit_default_formatting(self, subtests):
         ureg = UnitRegistry()
         x = ureg.Unit(UnitsContainer(meter=2, kilogram=1, second=-1))