SpatialPdfParser 1.0.5

SpatialPdfParser

<span style="font-size: 20px;">

SpatialPdfParser is a .NET library for structured data extraction from PDFs. It combines XML-based spatial mapping with Roslyn C# scripting to transform raw document coordinates into clean, validated data.

💡 Key Features

Spatial Extraction: Target specific regions of a PDF using X, Y coordinates and bounding boxes.

Dynamic Scripting: Use <% ... %> tags to execute real-time C# logic (splitting strings, regex, formatting) via Microsoft.CodeAnalysis.CSharp.Scripting.

Hierarchical Mapping: Define complex nested objects (e.g., Orders > Parties > Addresses) in a clean XML format.

Table Support: Specialized logic to segregate blocks of text into tabular data structures based on vertical proximity.

<div style="margin-top: 30px;">

🎯 Extraction Case Study: PO4.pdf**

</div>

Based on the this xml ,the po4.pdf (see image below) is scanned in spatial terms and the library maps raw text fragments to XML elements or attributes:

Header Data: Captures "PO Date" (2025-09-04) and "PO Number" (10346).

Party Identification: Segregates "Buyer" and "Supplier" details.

Complex Table Mapping: Iterates through the item table to create <line/> elements for items like Workstations.

<?xml version="1.0" encoding="utf-8"?>
<pdfMap client="IrisSystems" document="PurchaseOrder" rootName="po"    pdfSource="c:\\temp\\PO4.pdf">
 <po  number="&lt;%ScrapePDF(x:505,scanBelowY:793,width:50,line2LineGap:10) %&gt;" date="&lt;%ScrapePDF(x:168,scanBelowY:753,width:50,line2LineGap:10) %&gt;">
   <parties>
     <buyer  name="&lt;%ScrapePDF(x:46,scanBelowY:652,width:50,line2LineGap:10) %&gt;">
       <delivery date="&lt;%  ScrapePDF(x:427,scanBelowY:753,width:100,line2LineGap:10) %&gt;" />
       <address map="&lt;%  Split(ScrapePDF(x:46,scanBelowY:634,width:50,line2LineGap:10), new string[] {&quot;street&quot;, &quot;city&quot;, &quot;postcode&quot;, &quot;country&quot;}	) %&gt;"/>
       <contact map="&lt;% Split(ScrapePDF(x:46,scanBelowY:615,width:50,line2LineGap:10),new string[]{&quot;telephone&quot;,&quot;email&quot;},delimiter:'|') %&gt;"/>
     </buyer>
     <seller  name="&lt;%ScrapePDF(x:299,scanBelowY:653,width:50,line2LineGap:10) %&gt;">
         <address map="&lt;%  Split(ScrapePDF(x:298,scanBelowY:634,width:50,line2LineGap:10), new string[] {&quot;street&quot;, &quot;city&quot;, &quot;postcode&quot;, &quot;country&quot;}	) %&gt;"/>
       <contact map="&lt;% Split(ScrapePDF(x:298,scanBelowY:615,width:50,line2LineGap:10),new string[]{&quot;telephone&quot;,&quot;email&quot;},delimiter:'|') %&gt;"/>
       </seller>
   </parties>
   <po1loop map="&lt;%  SplitLinesToColumns(ScrapePDF(x:39,scanBelowY:502,width:515,line2LineGap:30),new string[]{ &quot;description&quot;,&quot;partnumber&quot;,&quot;qty&quot;,&quot;unitPrice&quot;,&quot;lineTotal&quot;}) %&gt;" />
 </po>
</pdfMap>

Input PDF Content

The source document contains structured tables for dates and items, along with blocks for buyer and supplier info., line items, and totals.

Generated XML Output

SpatialPdfParser transforms the raw text into the following structured format:

<pdfMap client="IrisSystems" document="PurchaseOrder" rootName="po" pdfSource="c:\\temp\\PO4.pdf">
 <po number="10346" date="2025-09-04">
   <parties>
     <buyer name="NextGen Tech Corp">
       <delivery date="2025-09-20"></delivery>
       <address street="77 Bay Street" city="Toronto" postcode="ON M5J 2L9" country="Canada"></address>
       <contact telephone="+1 (416) 555-7789" email="finance@nextgen.com"></contact>
     </buyer>
     <seller name="Metro Office Furniture Co.">
       <address street="1750 Market St" city="Denver" postcode="CO 80202" country="USA"></address>
       <contact telephone="+1 (303) 555-9922" email="sales@metrooffice.com"></contact>
     </seller>
   </parties>
   <po1loop>
     <line description="Workstations (Cubicle)" partnumber="MOF-WS600" qty="8" unitPrice="$ 850.00" lineTotal="6,800.00"></line>
     <line description="Storage Cabinets" partnumber="MOF-SC100" qty="6" unitPrice="$ 295.00" lineTotal="1,770.00"></line>
     <line description="Task Chairs" partnumber="MOF-TC210" qty="15" unitPrice="$ 175.00" lineTotal="2,625.00"></line>
     <line description="Breakroom Tables" partnumber="MOF-BT50" qty="4" unitPrice="$ 310.00" lineTotal="1,240.00"></line>
     <line description="Rambutan" partnumber="RAM-BUT1" qty="20" unitPrice="$ 40.00" lineTotal="800.00"></line>
   </po1loop>
 </po>
</pdfMap>

🔄 Execution Logic

The following diagram shows how SpatialPdfParser handles the po1loop in the XML : Code snippet

graph TD
    subgraph "PDF Layout Analysis"
        A[Identify Table Area] --> B[ScrapePDF at Coordinates]
        B --> C[Group Text by Y-Axis]
        C --> D[Sort Group by X-Axis]
    end
 
    subgraph "Scripted Parsing  Output"
        D --> E[Roslyn Script: SplitLinesToColumns]
        E --> F[Generate <line> nodes]
    end

📐 Coordinate Reference

SpatialPdfParser uses the standard PDF coordinate system where the origin (0,0) is at the Bottom-Left.

Note: If you are using coordinates from a design tool (like Figma or Chrome DevTools), you will likely need to invert your Y-axis values to match the PDF coordinate system.