Memelang v7

Memelang is an AI-optimized language for querying structured data, knowledge graphs, and retrieval-augmented generation pipelines.

Memes

• Comprises key-value pairs <keyopr><keyset><valopr><valset>
• Pairs separated by whitespaces \s+
• Terminates with semicolon ;

Pairs

• <keyopr> is either (none) or !
• <key> is alphanumeric string movie
• <keyset> is either
  • Single <key>
  • Single lone wildcard to match any key *
  • Variable (explained below)
  • CSV-style-comma-separated OR list movie,role,@1
• <valopr> is either = > < >= <= !=
• <value> is either
  • integer -123
  • or float 45.6
  • or unquoted alphanumeric string Leia
  • or CSV-style double quoted string "Anakin ""Ani"" Skywalker"
• <valset> is either
  • Single <value>
  • Single lone wildcard to match any value *
  • Variable (explained below)
  • CSV-style-comma-separated OR list Leia,"Anakin ""Ani"" Skywalker",@1

Others

• -> shorthand for m!=@m (below)
• Comments start with // and end with \n

Example Stored Memes

• Stored meme pairs are always certain <key>=<value>
• First pair is always m=<id>

m=100 actor="Mark Hamill" role="Luke Skywalker" movie="Star Wars" rating=4.5;
m=101 actor="Harrison Ford" role="Han Solo" movie="Star Wars" rating=4.6;
m=102 actor="Carrie Fisher" role=Leia movie="Star Wars" rating=4.2;

m=110 actor="Mark Hamill" role=Joker movie="Batman: Mask of the Phantasm" rating=4.7;
m=111 actor="Harrison Ford" role="Indiana Jones" movie="Raiders of the Lost Ark" rating=4.8;
m=112 actor="Carrie Fisher" role=Marie movie="When Harry Met Sally" rating=4.3;

m=200 person="Mark Hamill" birthyear=1951 birthplace="Oakland, CA";
m=201 person="Harrison Ford" birthyear=1942 birthplace="Chicago, IL";
m=202 person="Carrie Fisher" birthyear=1956 birthplace="Burbank, CA";

m=300 place="Oakland, CA" population=433000 climate=Mediterranean foundedyear=1852;
m=301 place="Chicago, IL" population=2740000 climate="Humid Continental" foundedyear=1833;
m=302 place="Burbank, CA" population=105000 climate=Mediterranean foundedyear=1887;

Example Query Memes

• At least one pair in a query contains uncertainty

// All movies with Mark Hamill as an actor
actor="Mark Hamill" movie=* rating>4 role=*;

// Response
m=100 actor="Mark Hamill" movie="Star Wars" rating=4.5 role="Luke Skywalker";
m=110 actor="Mark Hamill" movie="Batman: Mask of the Phantasm" rating=4.7 role=Joker;

// All keys and values from all memes relating to Mark Hamill
// *=* matches all pairs in the meme
*="Mark Hamill" *=*;

Example OR/Negation Query Memes

• ! keyopr negates all keys in the list
• != valopr negates all values in the list

// OR–list semantics in set notation
K1,K2=V1,V2;	// (key ∈ {K1,K2}) ∧ (value ∈ {V1,V2})
K1,K2!=V1,V2;	// (key ∈ {K1,K2}) ∧ (value ∉ {V1,V2})
!K1,K2=V1,V2;	// (key ∉ {K1,K2}) ∧ (value ∈ {V1,V2})
!K1,K2!=V1,V2;	// (key ∉ {K1,K2}) ∧ (value ∉ {V1,V2})
!K1,K2>V1;		// (key ∉ {K1,K2}) ∧ (value > V1)

// (actor OR role) = ("Luke Skywalker" OR "Mark Hamill")
actor,role="Luke Skywalker","Mark Hamill" movie=*;

// Actors who are not Mark Hamill or Carrie Fisher
actor!="Mark Hamill","Carrie Fisher" role=* movie=*;

// All keys, except actor and role, for Mark Hamill
!actor,role="Mark Hamill" movie=*;

// FAILURES
K1=V1=V2;		// Error: cannot chain values
K1=*K2=*K3=X;	// Error: missing spaces between pairs
K1=*,V1,V2;		// Error: cannot mix wildcard and commas
actor="*";		// Warning: wildcard inside quotes is literal "*"

Variables

Variables let later pairs retrieve the keys (list of strings) and values (list of integers/floats/strings) returned from prior pairs.

• Variables cannot be assigned.
• Variables cannot future-reference.
• Variables cannot reference objects or properties.
• Variables cannot be inside quotes.

Variable Arrays

Simplified variable model in python:

variables = {'K':[],'V':[]}

def var_add(keys: list, values: list):
	variables['K'].append(keys)
	variables['V'].append(values)
	if len(keys)==1: variables.setdefault(keys[0].lower(), []).append(values)

def index_get(n: str) -> int:
	if len(n)<1: return 1
	return int(n)

def var_get(tok):
	for sigil, mult, offset in (('#',1,-1), ('@',-1,0)):
		if tok.startswith(sigil):
			if tok.startswith(sigil+sigil):	return variables['K'][mult*index_get(tok[2:])+offset]
			if tok[1:].isdigit(): 			return variables['V'][mult*index_get(tok[1:])+offset]
			name, _, p = tok[1:].partition(':')
			return variables[name.lower()][mult*index_get(p)+offset]

While processing the query movie="Star Wars" role=* *>4.5, variable arrays grow as:

After pair	𝑝	variables
movie="Star Wars"	𝑝=1	{'K': [['movie']], 'V': [['Star Wars']], 'movie': [['Star Wars']]}
role=*	𝑝=2	{'K': [['movie'], ['role']], 'V': [['Star Wars'], [V1, V2, …]], 'movie': [['Star Wars']], 'role': [[V1, V2, …]]}
*>4.5;	𝑝=3	{'K': [['movie'], ['role'], [K1, K2, …]], 'V': [['Star Wars'], [V1, V2, …], [4.5]], 'movie': [['Star Wars']], 'role': [[V1, V2, …]]}

Forward-Indexes

• Sigils # and ## use absolute forward-indexing 𝑝
• Forward-index 𝑝 is a pair's absolute left-to-right query position
• Easy for AIs to count, hard for humans to maintain
• 𝑝≥1 and 𝑝 < current pair
• 𝑝=0 at ; or BOF
• 𝑝++ for every query pair (including m and -> pairs below)
• 𝑝=1 is the first pair
• 𝑝=2 is the second pair

Forward-Index Values

• #𝑝 retrieves the values from the 𝑝th pair
• #1 retrieves the values from the first pair
• #2 retrieves the values from the second pair

Forward-Index Keys

• ##𝑝 retrieves the keys from the 𝑝th pair
• ##1 retrieves the keys from the first pair
• ##2 retrieves the keys from the second pair

Forward-Index Keyname Values

• #keyname:𝑝 retrieves the values from the 𝑝th keyname pair
• #keyname and #keyname:1 retrieve the values from the first keyname pair
• Only populates when query keys is a single string (no * ! ,)
• Case-insensitive

Backward-Indexes

• Sigils @ and @@ use relative backward-indexing 𝑞
• Backward-index 𝑞 counts backward from the current pair's query position 𝑞 = 𝑝₂ − 𝑝₁
• Harder for AIs to count, easier for humans to maintain
• 𝑞≥1 and 𝑞≤𝑝-1
• 𝑞=0 is current pair (illegal)
• 𝑞++ for every query pair back (including m and -> pairs below)
• 𝑞=1 is the first pair back
• 𝑞=2 is the second pair back

Backward-Index Values

• @𝑞 retrieves the values from the 𝑞th pair back
• @1 retrieves the values from the first pair back
• @2 retrieves the values from the second pair back

Backward-Index Keys

• @@𝑞 retrieves the keys from the 𝑞th pair back
• @@1 retrieves the keys from the first pair back
• @@2 retrieves the keys from the second pair back

Backward-Index Keyname Values

• @keyname:𝑞 retrieves the values from the 𝑞th prior keyname pair
• @keyname and @keyname:1 retrieve the values from the last keyname pair
• Only populates when query keys is a single string (no * ! ,)
• Case-insensitive

Example Variables

Variables primarily retrieve prior pairs in joins (section below). However, variables occasionally retrieve pairs from the current meme (expert only, avoid).

// Titular roles whose value equals the movie title
// Every variable retrieves role=*
role=* actor=* movie=@role;
role=* actor=* movie=#role;
role=* actor=* movie=@2;
role=* actor=* movie=#1;
actor=* role=* movie=@role;
actor=* role=* movie=#role;
actor=* role=* movie=@1;
actor=* role=* movie=#2;

// Variable in comma list
role=* movie=@1,"Star Wars";

// Variable swap value into key name
K1=* @1=V2; // @1 must be alphanumeric
K1=* @K1=V2;

// Variable swap key name into value
*=V1 K2=@@1;
*=V1 K2=##1;

// FAILURE
actor="@person"; // Warning: variable inside quotes
actor=@person; // Likely meant

Joins

Distinct items (actors, movies, etc.) usually occupy distinct memes with unique m=id identifiers. By default, a pair stays within the current meme. A join query matches multiple memes by specifying a pair with key m, allowing the next pair to match a different meme.

• m key in a pair specifies which memes to join next
• @m is automatically populated with the current m=<id>
• m=@m and m=@m:1 stay in current meme (implicit default)
• m!=@m (shorthand ->) join to a different meme
• m=* join to any meme (current or different)

Join Variables

ALWAYS after an m pair, the following pair retrieves an uncertain pair from the prior meme.

• Preferred @keyname
  • keyname pair MUST be explict in prior meme
• Alternative @2 is usually the correct value variable for the prior meme's last pair
• ONLY join keys with semantically similar values
  • YES person=@author or birthyear=@foundedyear
  • NO person=@birthyear
• m!=@m joins to distinct memes, so trailing distinct conditionals like actor!=@actor are generally redundant

// The most common join is distinct
K1=*	// pair in first meme
m!=@m	// next pair must match a distinct meme
K2=@K1;	// pair in second meme, where K2 and K1 share a value

// Wildcard m allows the current meme in the second joined meme 
K1=*	// pair in first meme
m=*		// next pair may match any meme
K2=@K1;	// pair in second meme, where K2 and K1 share a value

// Wildcards in keys
*=*		// every pair in first meme
->		// (shorthand for m!=@m) next pair must match a distinct meme
*=@2;	// pair in second meme, any key has value from first meme

Example Joins

// All of Mark Hamill's costars - see how variables retrieve movie=* in their respective queries
// Each variable retrieves movie=*
actor="Mark Hamill" movie=* -> movie=@movie actor=*;
movie=* actor="Mark Hamill" -> movie=@movie actor=*;
actor="Mark Hamill" movie=* -> movie=@2 actor=*;
movie=* actor="Mark Hamill" -> movie=@3 actor=*;
actor="Mark Hamill" movie=* -> movie=#2 actor=*;
movie=* actor="Mark Hamill" -> movie=#1 actor=*;

// Response - join queries return combined memes, each pair belongs to the preceding m=<id>
m=100 actor="Mark Hamill" movie="Star Wars" m=101 movie="Star Wars" actor="Harrison Ford";
m=100 actor="Mark Hamill" movie="Star Wars" m=102 movie="Star Wars" actor="Carrie Fisher";

// People born in the year their birthplace was founded - join on int birthyear is year = foundedyear is year
person=* birthyear=* -> foundedyear=@birthyear place=*;
person=* birthyear=* -> foundedyear=@2 place=*;

// Every meme related to any value related to Mark Hamill - wild join keys (expert, avoid)
actor="Mark Hamill" *=* -> *=@2 *=*;

Example Multi-joins

// Other movies Mark Hamill's costars have acted in - multi join
actor="Mark Hamill" movie=* -> movie=@movie actor=* -> actor=@actor movie=*;
actor="Mark Hamill" movie=* -> movie=@2 actor=* -> actor=@2 movie=*;

// Population of the birthplace of the Star Wars cast
movie="Star Wars" actor=* -> person=@actor birthplace=* -> place=@birthplace population=*;
movie="Star Wars" actor=* -> person=@2 birthplace=* -> place=@2 population=*;

// Cities older than Burbank and with larger populations - inequality join
place="Burbank, CA" foundedyear=* population=* -> population>@population foundedyear<@foundedyear place=*;
place="Burbank, CA" foundedyear=* population=* -> population>@2 foundedyear<@4 place=*; // @4 = 1887 (Burbank foundedyear)

Failure Joins

movie=* movie=@1; // Warning: missing m= pair in join
movie=* -> movie=@2; // Likely meant

movie=* -> actor=*; // Warning: missing retrieved variable
movie=* -> movie=@2 actor=*; // Likely meant

movie=*->movie=@2; // Error: missing spaces around ->
movie=* -> movie=@2; // Likely meant

movie=* -> movie=@1; // Warning: wrong index - @1 retrieves -> pair
movie=* -> movie=@2; // Likely meant

birthplace=* person=* -> actor=@birthplace; // Warning: joined dissimilar values - actor != birthplace
person=* -> actor=@person birthplace=*; // Likely meant

movie=* -> actor=@director; // Semantic Error: undefined variable @director
director=* movie=* -> actor=@director; // Likely meant

director=* movie=* -> actor=@director:2; // Semantic Error: undefined variable @director:2 - wrong index
director=* movie=* -> actor=@director; // Likely meant

Credits

Bri Holt · Patents Pending · ©2025 HOLTWORK LLC