Problem and solution 

Software development in industry, building and many other fields requires working with CAD drawings. The most popular CAD formats are AutoCAD DWG and AutoCAD DXF, the latter being "simplified" dwg - a special format to be used by developers. The problem is that DXF and DWG formats are really complicated. They have dozens of objects with hundreds of interaction tricks and thousands of properties. Official DXF Reference from Autodesk has 256 pages though it fails to describe many important facts. Hence the development for CAD drawings is often required but is not easy to implement. This article is to tell you how to write the DXF reader in C#, what problems can arise and of course you can find example in C# source code, which is free for use under MPL license. 


DXF Structure 

DXF is an open ascii format from Autodesk and you can easily find documentation on it in the web. Here are some words about it. Below is a very simple example, to show the main parts:


0 - introduction of extended symbol names, following the "0"

SECTION, ENDSEC - begin / end of section. Sections can include Header, Entities, Objects. In the above code you see only Entities section where the entities are.

LINE - begins LINE entity description. Lines:



mean X1 double value. The value after 20 is Y1, after 30 - Z1 (0 in 2D drawings). 11, 21 and 31 codes are consequently for X2, Y2, Z2. So here we see the line with coordinates (39.19.., 36, 45.. - 39,19.., 736,45..) - this is vertical line.

So our aim is to read this acsii format. We need to load the file to stream and to take lines - even lines (0, 2, 4..) are CODE,  odd lines (1, 3, 5...) are VALUE and repeat this procedure step by step till the end of file "EOF". 

// take a pair of lines in DXF file, repeate this till "EOF":
public void Next()
  FCode = Convert.ToInt32(FStream.ReadLine()); //code
  FValue = FStream.ReadLine(); // value
// for code=0 we create entities. Entitites here are not only thouse which visible in AutoCAD.
// Entities can be also names of Sections and many other internal DXF objects.
// This method is called for all FCode == 0
public DXFEntity CreateEntity()
  DXFEntity E;
  switch (FValue)
    case "ENDSEC":
      return null;   // here we do not create entity
    case "ENDBLK":
      return null;
    case "ENDTAB":
      return null;
    case "LINE":    // for "LINE" value we create DXFLine object
      E = new DXFLine();
    case "SECTION": // "SECTION" will be object to store other objects like Line
      E = new DXFSection();
    case "BLOCK": // create block object
      E = new DXFBlock();
    case "INSERT": // insert is reference to block.
      E = new DXFInsert();
    case "TABLE":
      E = new DXFTable();
    case "CIRCLE":
      E = new DXFCircle();
    case "LAYER":
      E = new DXFLayer();
    case "TEXT":
      E = new DXFText();
    case "MTEXT":
      E = new DXFMText();
    case "ARC":
      E = new DXFArc();
    case "ELLIPSE":
      E = new DXFEllipse();
    default: // there are many other objects are possible. For them we create empty Entity
      E = new DXFEntity();
  // each Entity will need referense to the Base object Converter, which stores all Entities.
  E.Converter = this;
  return E; // return Entity and after it is added to the array of Entities

The method to read properties of entities is essentially similar to the one described above but it has one  important point: different entities have both identic and different properties. For instance many Entities have "base point" - in DXF, which is described by codes: 10 (x), 20 (y), 30(z):   


These codes are the same for LINE, CIRCLE, ELLIPSE, TEXT and for many others. So we can make Object-Oriented structure to read and to store the properties in order to avoid double-coding. We will store Layer and Base Point  in entity "DXFVisibleEntity" which will be ancestor for all visible entities. Look at the code to read these properties: 

//base class for all visible entities
public class DXFVisibleEntity : DXFEntity
//Base point (x, y, z) for all entities
  public DXFImport.SFPoint Point1 = new SFPoint();
  // virtual function ReadProperty() is overriden in all descendants of Entity to read
  //specific properties
  public override void ReadProperty()
  // for the different codes we read values
    switch (Converter.FCode)
        //read Layer
      case 8:
        layer = Converter.LayerByName(Converter.FValue);
        //read Coordinates
      case 10: //X
        Point1.X = Convert.ToSingle(Converter.FValue, Converter.N);
      case 20: //Y
        Point1.Y = Convert.ToSingle(Converter.FValue, Converter.N);
        //read Color
      case 62:
        FColor = CADImage.IntToColor(Convert.ToInt32(Converter.FValue, Converter.N));

We use the same approach to read the second coordinate in LINE, radius in Circle and so on.  

DXF File and DXF Import .NET Structure   


This scheme shows main parts of DXF file and the way they are connected with the C# source code in the project. The dash lines stand for associations between DXF file objects and objects, programmed in C#   

CADImage is a class for loading from DXF file and drawing to Graphics. It stores the DXF Entities in field:

public DXFSection FEntities;  

In the scheme it is DXFSection.

DXFEntity is base class for all Entities classes. Classes DXFBlocks and DXFSection are not visible. Class DXFVisibleEntity is the ancestor for all visible Entities.

By the way, the scheme above is made in DXF format:) in ABViewer software.

CAD tricks

If you are not familiar with AutoCAD please pay attention to the structure of DXF entities . There is a special entity "Block" which may have many "inserts" in the CAD drawing. Block is just set of entities (including nested blocks) which can be inserted many times.


 Block changes many properties of element when showing it. So if you want to know the color of entity, it is not enough to read  "Entity.Color"  - it is necessary to see all the Inserts and Blocks, in which this entity can be included. To get the correct color in DXF Import .NET project we made the following function: EntColor(DXFEntity E, DXFInsert Ins).

Please use EntColor() function get the correct Color type even if you do not have Blocks in the file. Pay attention that the most common color is "ByLayer" and in oder to read the correct color we need to  read the color from Layer entity.  This functionality is also provided in this function.

Below is EntColor function. It has many tricks, for instance checking the == "0" - in DXF layer "0" is special and elements with color "ByLayer" get the color from Block if they are on "0" layer.

<pre lang="cs">//Use this func to know the color of Entity, DXFInsert is Insert entity or null. 
public static Color EntColor(DXFEntity E, DXFInsert Ins)
  DXFInsert vIns = Ins;
  DXFEntity Ent = E;
  Color Result = DXFConst.clNone;
  if(Ent is DXFVisibleEntity) Result = E.FColor;
  /*if(Ent is Polyline)
    Result = ((Polyline)Ent).Pen.Pen.Color;*/
  if(E.layer == null) return  Result; 
/* algorithm is rather difficult here. This is the way, how AutoCAD works with the colors,
if you try to create entities in AutoCAD, you will see how they use Colors */ 
  if((Result ==  clByLayer)||(Result == clByBlock))
    if((vIns == null)||((Result == clByLayer)&&( != "0")))
      if(Result == clByLayer)
        if(Ent.layer.color != clNone)
          Result = Ent.layer.color;
        else Result = Color.Black;
      while(vIns != null)
        Result = vIns.color;
        if((Result !=  clByBlock) && !((Result ==  clByLayer) &&
          ( == "0")))
          if(Result ==  clByLayer)
            Result = vIns.layer.color;
        if((vIns.owner == null)&&(Result == clByLayer))
          Result = vIns.layer.color;
        vIns = vIns.owner;
  if((Result == clByLayer)||(Result == clByBlock))
    Result = clNone;
  return Result;

How to use the software 

Main code is in DXFImport.cs. You can just use this file in your project or see to it as an example of reading and visualization of DXF files.  

Sample code to use DXFImport.cs is Form1.cs. 

How to view entities 

In Form1.cs we use the Form1_Paint event: 

private void Form1_Paint(object sender, System.Windows.Forms.PaintEventArgs e)
 //FCADImage - base class to reference to DXF
  if (FCADImage == null)
  FCADImage.Draw(e.Graphics); // CADImage.Draw() accepts Graphics to draw to

We can use FCADImage.Draw() for drawing to any Graphics - for instance to printer.  FCADImage.Draw() function should use a special algorithm, when each entity is drawn with the use of block/insert/layer parameters.   

public void Draw(Graphics e)
  if (FMain == null)
  FGraphics = e;
  // Iterate begins to  over the entities to draw all of them
  FEntities.Iterate(new CADEntityProc(DrawEntity), FParams);

Pay attention to FEntities.Iterate() func, it allows accessing all entities including their being inside the blocks.  This is how it works: There is a class DXFGroup which is an ancestor of Entity and which can store array of Entities. For instance, Block is ancestor of DXFGroup and can store many entities like LINE. For better unification we can use ONE BASE GROUP ENTITY object for all the drawing, this Entity will have array of all entities, each of them can also be the Group - the classic "Tree". 

Iterate method finally should call Draw() for each Entity in oder to draw it to the Graphics:

<span style="white-space: pre; " class="Apple-tab-span">    </span>protected static void DrawEntity(DXFEntity Ent)
<span style="white-space: pre; " class="Apple-tab-span">        </span>{
<span style="white-space: pre; " class="Apple-tab-span">            </span>Ent.Draw(FGraphics);
<span style="white-space: pre; " class="Apple-tab-span">        </span>} 

Draw() method is overridden in descendants of Entity to draw particular entities. Let us see in details how it is implemented in DXFLine: 

// draw line
public override void Draw(System.Drawing.Graphics G)
    // points P1  (x1, y1) and P2 (x2, y2) of Line
    SFPoint P1, P2;
    // read color via EntColor to get real color
    Color RealColor = DXFConst.EntColor(this, Converter.FParams.Insert);
    // Read point 1 -convert global coordinates to the screen coordinates:
    P1 = Converter.GetPoint(Point1);
    //read point 2
    P2 = Converter.GetPoint(Point2);
    if (FVisible)
      G.DrawLine(new Pen(RealColor, 1),  P1.X, P1.Y, P2.X, P2.Y);

1. We get the Real Color via DXFConst.EntColor(this, Converter.FParams.Insert); as I described before.

2. Points are converted from Global Coordinates to screen coordinates in function GetPoint(). GetPoint not only converts global-to-screen but also uses Block offsets and block scale inside. Thus it facilitates the development work, eliminating the need to "see" what block is being drawn at the moment - Block changes "FParams.matrix" to draw itself. And all entities coordinates use this "FParams


3. Entity is drawn to the given Graphics G: 

So you can draw to printer, to raster image or to other Graphics.

DXF Import .NET Reference:

public class DXFConst 

Stores constants and base functions.

public class DXFMatrix 

Class to work with coordinates.

public struct FRect 

Description of 3D space where the CAD drawing is situated in global coordinates.

public struct CADIterate

Stores all needed parameters for entities when processing "Iterate" function.

public class CADImage 

Class to draw CAD drawing.

public class DXFEntity 

Base class for all DXF Entities

public class DXFGroup : DXFEntity

Base class for all group entities.

public class DXFTable : DXFGroup 

Class to read from DXF "Table" section - here it reads only Layers.

public class DXFVisibleEntity : DXFEntity

Base class for all visible entities (unvisible - are "DXFTable", "DXFLayer", etc.)

public class DXFCustomVertex: DXFVisibleEntity 

Special class for 3D point in DXF.

public class DXFText: DXFCustomVertex

Stores and Draws "Text" DXF entity.

the following classes are for particular DXF entities:

public class DXFLine : DXFVisibleEntity
public class DXFArc: DXFCircle
public class DXFEllipse: DXFArc
public class DXFLayer: DXFEntity
public class DXFBlock : DXFGroup

Class to work with DXF Block. 

public class DXFInsert : DXFVisibleEntity  

Class to work with "Insert" in DXF, for AutoCAD users this is "Block reference". Blocks are not visible, Inserts are visible. 


The purpose of the article is to give some advice how to write DXF readers. DXF file structure is  not so difficult as its logical  presentation. The article looks through the base DXF format problems and shows how to find solution for them. The example source is written in C# and may be helpful for all who need to have access to DXF files. 

Other projects on the Code Project

On The Code Project you can already find another DXF reader:

Main advantages of our project as compared to the project above are:

1. Block / Inserts

2. Layers

3. Text

which are not presented in that DXF reader.


This is the first open source version of  DXF Import .NET.

推荐.NET配套的通用数据层ORM框架:CYQ.Data 通用数据层框架