ASCII art diagram converter
Given the RFC 1035 message diagram from Section 4.1.1 (Header section format) as a string: http://www.ietf.org/rfc/rfc1035.txt
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ID | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ |QR| Opcode |AA|TC|RD|RA| Z | RCODE | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | QDCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ANCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | NSCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ARCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
Where (every column of the table is 1 bit):
ID is 16 bits QR = Query (0) or Response (1) Opcode = Four bits defining kind of query: 0: a standard query (QUERY) 1: an inverse query (IQUERY) 2: a server status request (STATUS) 3-15: reserved for future use AA = Authoritative Answer bit TC = Truncation bit RD = Recursion Desired bit RA = Recursion Available bit Z = Reserved RCODE = Response code QC = Question Count ANC = Answer Count AUC = Authority Count ADC = Additional Count
Write a function, member function, class or template that accepts a similar multi-line string as input to define a data structure or something else able to decode or store a header with that specified bit structure.
If your language has macros, introspection, code generation, or powerful enough templates, then accept such string at compile-time to define the header data structure statically.
Such "Header" function or template should accept a table with 8, 16, 32 or 64 columns, and any number of rows. For simplicity the only allowed symbols to define the table are + - | (plus, minus, pipe), and whitespace. Lines of the input string composed just of whitespace should be ignored. Leading and trailing whitespace in the input string should be ignored, as well as before and after each table row. The box for each bit of the diagram takes four chars "+--+". The code should perform a little of validation of the input string, but for brevity a full validation is not required.
Bonus: perform a thoroughly validation of the input string.
D
This solution generates anonymous struct code at compile-time, that can be mixed-in inside a struct or class. <lang d>string makeStructFromDiagram(in string rawDiagram) pure @safe {
import std.conv: text, format; import std.string: strip, splitLines, indexOf; import std.array: empty, popFront;
static void commitCurrent(ref uint anonCount, ref uint totalBits, ref uint currentBits, ref string code, ref string currentName) pure @safe { if (currentBits) { code ~= "\t";
currentName = currentName.strip; if (currentName.empty) { anonCount++; currentName = "anonymous_field_" ~ anonCount.text; }
string type; if (currentBits == 1) type = "bool"; else if (currentBits <= ubyte.sizeof * 8) type = "ubyte"; else if (currentBits <= ushort.sizeof * 8) type = "ushort"; else if (currentBits <= uint.sizeof * 8) type = "uint"; else if (currentBits <= ulong.sizeof * 8) type = "ulong"; //else if (currentBits <= ucent.sizeof * 8) // type = "ucent"; else assert(0, "Too many bits for the item "~ currentName);
immutable byteOffset = totalBits / 8; immutable bitOffset = totalBits % 8;
// Getter: code ~= "@property " ~ type ~ " " ~ currentName ~ "() const pure nothrow @safe {\n"; code ~= "\t\t"; if (currentBits == 1) { code ~= format("return (_payload[%d] & (1 << (7-%d)))"~ " ? true : false;", byteOffset, bitOffset); } else if (currentBits < 8) { auto mask = (1 << currentBits) - 1; mask <<= 7 - bitOffset - currentBits + 1; code ~= format("return (_payload[%d] & 0b%08b) >> %d;", byteOffset, mask, 7 - bitOffset - currentBits + 1); } else { assert(currentBits % 8 == 0); assert(bitOffset == 0); code ~= type ~ " v = 0;\n\t\t";
code ~= "version(LittleEndian) {\n\t\t"; foreach (immutable i; 0 .. currentBits / 8) code ~= "\tv |= (cast(" ~ type ~ ") _payload[" ~ text(byteOffset + i) ~ "]) << (" ~ text((currentBits / 8) - i - 1) ~ " * 8);\n\t\t"; code ~= "} else static assert(0);\n\t\t"; code ~= "return v;"; } code ~= "\n"; code ~= "\t}\n\t";
// Setter: code ~= "@property void " ~ currentName ~ "(in " ~ type ~ " value) pure nothrow @safe {\n"; code ~= "\t\t"; if (currentBits < 8) { auto mask = (1 << currentBits) - 1; mask <<= 7 - bitOffset - currentBits + 1; code ~= format("_payload[%d] &= ~0b%08b;\n\t\t", byteOffset, mask); code ~= "assert(value < " ~ text(1 << currentBits) ~ ");\n\t\t"; code~=format("_payload[%d] |= cast(ubyte) value << %d;", byteOffset, 7 - bitOffset - currentBits + 1); } else { assert(currentBits % 8 == 0); assert(bitOffset == 0);
code ~= "version(LittleEndian) {\n\t\t"; foreach (immutable i; 0 .. currentBits / 8) code ~= "\t_payload[" ~ text(byteOffset + i) ~ "] = (value >> (" ~ text((currentBits / 8) - i - 1) ~ " * 8) & 0xff);\n\t\t"; code ~= "} else static assert(0);"; }
code ~= "\n"; code ~= "\t}\n"; totalBits += currentBits; }
currentBits = 0; currentName = null; }
enum C : char { pipe='|', cross='+' } enum cWidth = 3; // Width of a bit cell in the table. immutable diagram = rawDiagram.strip;
uint bitCountPerRow = 0, anonCount = 0, totalBits, currentBits; string currentName; string code = "struct {\n"; // Anonymous.
foreach (line; diagram.splitLines) { line = line.strip; if (line[0] == C.cross) { commitCurrent(anonCount, totalBits, currentBits, code, currentName); if (bitCountPerRow == 0) bitCountPerRow = cast(uint) (line.length - 1) / cWidth; else assert(bitCountPerRow == (line.length - 1) / cWidth); } else { // A field of some sort. while (line.length > 2) { assert(line[0] != '/', "Variable length data not supported"); assert(line[0] == C.pipe, "Malformed table"); line.popFront; const idx = line[0 .. $ - 1].indexOf(C.pipe); if (idx != -1) { const field = line[0 .. idx]; line = line[idx .. $];
commitCurrent(anonCount, totalBits, currentBits, code, currentName); currentName = field; currentBits = (field.length + 1) / cWidth; commitCurrent(anonCount, totalBits, currentBits, code, currentName); } else { // The full row or a continuation of the last. currentName ~= line[0 .. $ - 1]; // At this point, line does not include the first // C.pipe, but the length will include the last. currentBits += cast(uint) line.length / cWidth;
line = line[$ .. $]; } } } }
// Using bytes to avoid endianness issues. // hopefully the compiler will optimize it, otherwise // maybe we could specialize the properties more. code ~= "\n\tprivate ubyte[" ~ text((totalBits + 7) / 8) ~ "] _payload;\n";
return code ~ "}";
}
void main() { // Testing.
import std.stdio;
enum diagram = " +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ID | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ |QR| Opcode |AA|TC|RD|RA| Z | RCODE | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | QDCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ANCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | NSCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ARCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+";
// To debug the code generation: //pragma(msg, diagram.makeStructFromDiagram);
// Usage. static struct Header { mixin(diagram.makeStructFromDiagram); }
Header h; h.ID = 10; h.RA = true; h.ARCOUNT = 255; h.Opcode = 7;
// See the byte representation to test the setter's details. h._payload.writeln;
// Test the getters: assert(h.ID == 10); assert(h.RA == true); assert(h.ARCOUNT == 255); assert(h.Opcode == 7);
}</lang>
- Output:
[0, 10, 56, 128, 0, 0, 0, 0, 0, 0, 0, 255]
Uncommenting the pragma(msg) in the main function, you can see that it generates code like: <lang d>struct {
@property ushort ID() const pure nothrow @safe { ushort v = 0; version(LittleEndian) { v |= (cast(ushort) _payload[0]) << (1 * 8); v |= (cast(ushort) _payload[1]) << (0 * 8); } else static assert(0); return v; } @property void ID(in ushort value) pure nothrow @safe { version(LittleEndian) { _payload[0] = (value >> (1 * 8) & 0xff); _payload[1] = (value >> (0 * 8) & 0xff); } else static assert(0); }
...
private ubyte[12] _payload;
}</lang>
It also supports larger values like this, that is 32 bits long:
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ThirtyTwo | | | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
J
<lang J>require'strings'
soul=: -. {. normalize=: [:soul' ',dltb;._2
mask=: 0: _1} '+' = {. partition=: '|' = mask #"1 soul labels=: ;@(([: <@}: <@dltb;._1)"1~ '|'&=)@soul names=: ;:^:(0 = L.)
unpacker=:1 :0
p=. , partition normalize m p #.;.1 (8#2) ,@:#: ]
)
packer=:1 :0
w=. -#;.1 ,partition normalize m _8 (#.\ ;) w ({. #:)&.> ]
)
getter=:1 :0
nm=. labels normalize m (nm i. names@[) { ]
)
setter=:1 :0
q=. ' n=. q,~q,;:inv labels normalize m 1 :('(',n,' i.&names m)}')
)
starter=:1 :0
0"0 labels normalize m
)</lang>
Sample definition (note the deliberate introduction of extraneous whitespace in locations the task requires us to ignore it.
<lang j>sample=: 0 :0
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
| ID | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ |QR| Opcode |AA|TC|RD|RA| Z | RCODE | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | QDCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ANCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | NSCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | ARCOUNT | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
)
unpack=: sample unpacker pack=: sample packer get=: sample getter set=: sample setter start=: sample starter</lang>
Example data for sample definition:
<lang J>
4095 13 5 6144 4096 'ID Opcode RCODE ARCOUNT QDCOUNT' set start
4095 0 13 0 0 0 0 0 5 4096 0 0 6144
pack 4095 13 5 6144 4096 'ID Opcode RCODE ARCOUNT QDCOUNT' set start
15 255 104 5 16 0 0 0 0 0 24 0
unpack 0 10 56 128 0 0 0 0 0 0 0 255
10 0 7 0 0 0 1 0 0 0 0 0 255
'Opcode' get unpack 0 10 56 128 0 0 0 0 0 0 0 255
7</lang>
In other words:
- unpack converts an octet sequence to the corresponding numeric sequence
- pack converts a numeric sequence to the corresponding octet sequence
- get extracts named elements from the numeric sequence
- set updates named elements in the numeric sequence
- start represents the default "all zeros" sequence which may be used to derive other sequences
Note that this implementation assumes that the ascii diagram represents the native word width on a single line, and assumes well formed data.