Parsing a text file containing device information

Question

After trying a couple of parsers, such as Log Parser 2.2, I ended up writing a small utility that supposedly parses the following information onto an object:

(from DirectX Caps Viewer, truncated)

DirectX Graphics Adapters
 AMD Radeon HD 7800 Series
 Driver aticfx32.dll
 Description AMD Radeon HD 7800 Series
 DriverVersion 524305656590
 VendorId 4,098
 DeviceId 26,649
 SubSysId -501,147,829
 Revision 0
 DeviceIdentifier {D7B71EE2-2B59-11CF-3570-2BC2BEC2C535}
 WHQLLevel 1
 Display Modes
 640 x 480 D3DFMT_X8R8G8B8 60
 1920 x 1200 D3DFMT_R5G6B5 60
 D3D Device Types
 HAL
 Caps
 DeviceType 1
 MaxPixelShaderValue 3.40282E+038
 Caps
 D3DCAPS_READ_SCANLINE Yes
 Caps2
 D3DCAPS2_CANCALIBRATEGAMMA No
 D3DCAPS2_DYNAMICTEXTURES Yes
 Caps3
 D3DCAPS3_ALPHA_FULLSCREEN_FLIP_OR_DISCARD Yes
 PresentationIntervals
 D3DPRESENT_INTERVAL_ONE Yes
 D3DPRESENT_INTERVAL_TWO Yes
 D3DPRESENT_INTERVAL_THREE Yes
 D3DPRESENT_INTERVAL_FOUR Yes
 D3DPRESENT_INTERVAL_IMMEDIATE Yes
 CursorCaps
 D3DCURSORCAPS_COLOR Yes
 D3DCURSORCAPS_LOWRES No
 Adapter Formats
 D3DFMT_X8R8G8B8 (Fullscreen)
 Back Buffer Formats
 D3DFMT_A8R8G8B8
 MultiSample Types
 D3DMULTISAMPLE_NONE
 D3DMULTISAMPLE_2_SAMPLES

Here's the code. It basically uses a regular expression to differentiate whether each line is an entry or a category. Then, using list, it will add that current item to its parent.

internal class Program
{
 private static void Main(string[] args)
 {
 using (var reader = new StreamReader(File.OpenRead(@"..\..\hd7850all.log")))
 {
 var categories =
 new List<Tuple<int, Category>>(new[] {new Tuple<int, Category>(-1, new Category {Name = "root"})});
 while (!reader.EndOfStream)
 {
 string s = reader.ReadLine();
 if (s == null) throw new InvalidOperationException("Input string cannot be null");
 //bool isCategory = Regex.IsMatch(s, @"^[\w]");
 //bool isSubCategory = Regex.IsMatch(s, @"^ +(?!.* {2,})");
 bool isAnyCategory = Regex.IsMatch(s, @"^ *(?!.* {2,})");
 bool isEntry = Regex.IsMatch(s, @"(?<!^ *) {2,}");
 int depth = Regex.Match(s, "^ *").Length;
 string name = Regex.Replace(s, "^ *", string.Empty);
 var category = new Category {Name = name};
 var value = new Tuple<int, Category>(depth, category);
 if (isAnyCategory)
 {
 for (int i = categories.Count - 1; i >= 0; i--)
 {
 Tuple<int, Category> tuple = categories[i];
 if (tuple.Item1 < depth)
 {
 tuple.Item2.Categories.Add(category);
 break;
 }
 }
 categories.Add(value);
 }
 else if (isEntry)
 {
 Tuple<int, Category> tuple = categories.Last();
 tuple.Item2.Entries.Add(new Entry {Name = name});
 }
 }
 Category root = categories[0].Item2;
 }
 }
}
public class Category
{
 public Category()
 {
 Categories = new List<Category>();
 Entries = new List<Entry>();
 }
 public string Name { get; set; }
 public List<Entry> Entries { get; set; }
 public List<Category> Categories { get; set; }
 public override string ToString()
 {
 return Name;
 }
}
public class Entry
{
 public string Name { get; set; }
 public override string ToString()
 {
 return Name;
 }
}

Do you know of a more efficient method or approach to transform such data to an object?

• Jamal: Yes, this question is on-topic.
• svick: Why are you looking for a more efficient approach? Is your code too slow? If it is, what did profiling tell you? What is the slowest part?
• aybe: Not really, it does work pretty well though some categories are seen as entries instead. It's more about getting a different point of view. And well, what you see here took me quite some time to get to it as, as usual I started with a complex solution and after many iterations I'm satisfied with it and it's really simple now. In short, I am expecting a review about it !

Answer 1

Your usage of whitespaces in regex patterns threw me off, I think they should be captured with \s.

Also your object model doesn't seem to reflect the file's format. I see this:

• GraphicAdapter

  • .Driver (="aticfx32.dll")
  • .Description (="AMD Radeon HD 7800 Series")
  • .DriverVersion (="524305656590")
  • .VendorId (="4,098")
  • .DeviceId (="26,649")
  • .SubSysId (="-501,147,829")
  • .Revision (="0")
  • .DeviceIdentifier (="{D7B71EE2-2B59-11CF-3570-2BC2BEC2C535}")
  • .WHQLLevel (="1")
  • .DisplayModes
  • .DeviceTypes

• DisplayMode

  • .Resolution (="640 x 480")
  • .WhateverThatIs (="D3DFMT_X8R8G8B8")
  • .RefreshRateHz (="60")

• DeviceType

  • .Identifier (="HAL")
  • .Caps
  • .AdapterFormats

And so on and so forth - what I mean is that there's much more than "entry" and "category" going on here, by merely capturing "entry" and "category" you're missing out on the richness of the data you're parsing.

• svick: While strongly typed access to data (similar to types generated from XML schema, if this was XML) is nice, I think there is a place for a more weak approach too (more akin to plain LINQ to XML).