[WIP]: Analysis Input - Simulates Rendered Text

Strips out link and image URLs, table markup, and other non-visible items that distract the analysis, giving false results. For example, URLs can be a source of many completion suggestions for numbers.

src/VirtualStash-Core/MailMessage.extension.st

@@ -0,0 +1,9 @@
+Extension { #name : #MailMessage }
+
+{ #category : #'*VirtualStash-Core' }
+MailMessage >> vsAnalysisInput [
+
+	^ self bodyRlHtml 
+		ifNotNil: [ :html | html vsAnalysisInput ]
+		ifNil: [ self bodyTextFormatted ]
+]

src/VirtualStash-Core/RlEmail.extension.st

@@ -23,6 +23,12 @@ RlEmail >> possibleCurrencyAmounts [
 	^ matches asSet
 ]
 
+{ #category : #'*VirtualStash-Core' }
+RlEmail >> vsAnalysisInput [
+
+	^ self mailMessage vsAnalysisInput
+]
+
 { #category : #'*VirtualStash-Core' }
 RlEmail >> vsCounterpartyGuess [
 

src/VirtualStash-Core/RlHTML.extension.st

@@ -0,0 +1,19 @@
+Extension { #name : #RlHTML }
+
+{ #category : #'*VirtualStash-Core' }
+RlHTML >> vsAnalysisInput [
+	"Return 'just the plain text the user would see if rendered'"
+
+	^ PBApplication uniqueInstance newCommandStringFactory
+		bindingAt: #htmlString put: self contents;
+		script: 'import html2text
+text_maker = html2text.HTML2Text()
+text_maker.ignore_links = True
+text_maker.ignore_images = True
+text_maker.ignore_tables = True
+text_maker.ignore_emphasis = True';
+		resultExpression: 'text_maker.handle(htmlString)';
+		sendAndWait	
+
+	"Reference: https://github.com/Alir3z4/html2text/blob/master/docs/usage.md#using-options"
+]