diff --git a/bin/choose-math-char b/bin/choose-math-char
new file mode 100755
index 0000000..bc34650
--- /dev/null
+++ b/bin/choose-math-char
@@ -0,0 +1,47 @@
+#!/usr/bin/env osh
+
+# Create an interface for select Unicode characters with rofi and copy it into the clipboard.
+# The characters are extracted from the Unicode Character Database (https://www.unicode.org/ucd/).
+
+# Dependencies:
+# * Oil shell
+# * curl
+# * python (at least v3.7)
+# * rofi
+# * awk
+
+# Set to use Oil features and strictness.
+shopt --set strict:all oil:all
+
+var UCD_VERSION = "13.0.0"
+var UCD_XML_URL = "https://www.unicode.org/Public/$UCD_VERSION/ucdxml/ucd.nounihan.grouped.zip"
+var CACHE = ${XDG_CACHE_DIR:-$HOME/.cache/unicode-character-database}
+mkdir -p $CACHE
+
+if test ! -f $CACHE/ucd.zip {
+  curl $UCD_XML_URL --output $CACHE/ucd.zip --silent
+}
+
+# The remaining thing is coded with Python because I want to use the UCD with no Unihan data and the grouped variation.
+# Compared to the other variations, it is only ~6MB compared to ~55MB for the flat variation.
+# Also, it requires more conditional handling than a simple shell script at this point.
+# I could've made this script entirely in Python but I want to see what Oil shell is capable of.
+python <<CODE | rofi -dmenu -p "Choose mathematical character" | awk '{ print $1 }' | xclip -selection clipboard
+import xml.etree.ElementTree as ET
+
+root = ET.fromstring('''$(unzip -p $CACHE/ucd.zip ucd.nounihan.grouped.xml)''')
+
+# Print '<char>' and recurse into '<group>'.
+def print_char(element):
+  if element.tag == '{http://www.unicode.org/ns/2003/ucd/1.0}char':
+    alias = element.get('na') if element.get('na') else element.get('na1')
+    codepoint = int(element.get('cp'), 16)
+    print("{code} {alias}".format(code=chr(codepoint), alias=alias))
+  elif element.tag == '{http://www.unicode.org/ns/2003/ucd/1.0}group':
+    for child in list(element):
+      print_char(child)
+
+math_nodes = root.findall('./{http://www.unicode.org/ns/2003/ucd/1.0}repertoire//*[@Math="Y"]')
+for point in math_nodes:
+  print_char(point)
+CODE
diff --git a/bin/generate-gitignore b/bin/generate-gitignore
index cb51995..d2d7fce 100755
--- a/bin/generate-gitignore
+++ b/bin/generate-gitignore
@@ -18,7 +18,7 @@ CACHE_FILE="${XDG_CACHE_DIR:-$HOME/.cache}/gitignore-io.langs.json"
 # Check if the language list is downloaded for the last hour (3600 seconds).
 if [ ! -e $CACHE_FILE ] || test $(expr $(date "+%s") - $(date -r $CACHE_FILE "+%s")) -gt 3600
 then
-    curl --silent --location --output $CACHE_FILE "https://gitignore.io/api/list?format=json"
+    ping "gitignore.io" --count 4 && curl --silent --location --output $CACHE_FILE "https://gitignore.io/api/list?format=json"
 fi
 
 KEYS=$(jq 'keys | .[] | @text' --raw-output $CACHE_FILE | fzf --multi | while read lang; do echo " .[\"$lang\"].contents"; done | paste -s -d ',')