Eval#
Introduction#
The eval module is a general purpose filtering and programming environment that runs a “program” against each entry in the query pipeline. It uses a C-style syntax and supports creating simple filters, as well as arbitrary operations on each entry.
For example, to filter entries to just those where the enumerated value foo is equal to bar or baz:
tag=gravwell json foo | eval ( foo == "bar" || foo == "baz" )
Additionally, C-style statements can be used for more advanced operations. For example, to conditionally create an EV:
tag=gravwell json foo | eval if (foo == "bar") { other_variable = "hey! foo is bar!"; }
The rest of this document describes the syntax and semantics of the eval language.
Use cases#
Eval is used primarily for advanced filtering, especially multifiltering. For example, in order to filter a field extracted from the json module, you would normally use filtering in the json module directly:
tag=gravwell json foo=="my value"
However, if you want to filter “foo” in the example above to “my value” or “my other value”, you must use the eval module:
tag=gravwell json foo | eval ( foo == "my value" || foo == "my other value" )
Lexical elements#
Identifiers#
Identifiers name variables and types. For example, in foo = "bar";, “foo” is an identifier. Identifiers can contain any number of letters, digits, and the underscore “_” character, and must begin with either a letter or an underscore. No other characters are allowed.
It is possible to create an enumerated value that doesn’t follow the identifier syntax rules in eval. When using enumerated values like this, you can wrap the name of the enumerated value in $( ). For example:
tag=gravwell json "55 crazy enumerated value!" | eval $(55 crazy enumerated value!) == "value"
For example, the following are all valid identifiers.
foo
_foo
foo314
foo_bar
The identifiers “true” and “false” are always predeclared, and are reserved words.
Variables#
Variables are enumerated values and when declared can be used by other modules in the query pipeline. For example:
tag=gravwell eval foo = "bar"; | table foo
Additionally, all previously produced enumerated values in the query pipeline are available as typed variables. For example:
tag=gravwell json foo | eval foo == "bar"
Persistent Variables#
Normally, eval executes against every entry in the query, with no state carried between entries. Eval has support for “persistent” variables, that persist throughout the execution of a query. For example, to use eval to count the number of entries in a query:
tag=gravwell
eval
var count = 0;
count++;
output = count;
| table
This program will initialize a variable “count” to 0, and the value will persist across entries.
To use a persistent variable, it must be declared with var <variable name>;. Optionally, you can initialize the variable to a value with the syntax var <variable name> = <expression>;
Persistent variables are not attached to entries like other variables. In order to use a persistent variable’s value outside of eval, it must be assigned to a regular variable.
Persistent Maps#
Like persistent variables, eval can also create persistent maps, which behave like key/value objects. A map uses strings as keys, and can store any eval variable type except maps. Maps are declared with the map keyword, and accessed like other variables. To access a specific key in a map, the notation map[key] is used.
For example, to count each unique Appname in a list of syslog entries, a map can be used with the syslog Appname as the key:
tag=gravwell
syslog Appname
| eval
map appnames;
if (Appname == "")
appnames[Appname] = 0; // The key doesn't exist. Create one.
else
appnames[Appname]=appnames[Appname]+1; // The key does exist. Increment.
Like persistent variables, maps are not attached to entries. Values must be assigned to regular variables in order to use them outside of eval.
Maps have a limit of 1000000 keys. Any new key assigned to a map after this limit is reached will be discarded.
Keywords#
The following keywords are reserved and may not be used as identifiers.
if
else
for
break
continue
return
(the remaining keywords are not currently supported, but are reserved)
case
default
switch
Operators#
Operators are used for filtering and boolean/arithmetic logic. Operators behave similarly to all C-style programming languages (C, C++, Go, …) with one caveat – the bitwise OR operator | can only appear inside parenthetical expressions in order to avoid a syntax conflict with the Gravwell query language.
The following operators are supported.
+ - * / Arithmetic operators
% Modulo
& | Bitwise operators
<< >> Shift operators
&& || ! Boolean operators
-- ++ Postfix operators
< > <= >= == != Equality operators
~ !~ String "contains" operators
Punctuation#
Eval uses punctuation similar to the C programming language. For example, ( ) are used to enclose expressions in statements as in if (foo == "bar")....
The following character sequences are reserved punctuation.
( ) Expression and function grouping
{ } Statement and block grouping
, Function parameter separator
; Statement separator
$ Alternative enumerated value accessor
[ ] Array indexing (not currently supported)
. : Scope and range (not currently supported)
Numbers#
Numbers can be expressed as decimal and hexadecimal integers, as well as decimal floating-point. Hexadecimal values always begin with “0x”.
The following examples are valid numbers.
5
-5
0x05
5.00
String literals#
Strings are any sequence of characters enclosed in double quotes ".
Variables and types#
All enumerated values created in the query pipeline are available in eval as variables. Additionally, all variables created in eval are available as enumerated values to any modules after eval in the query pipeline.
Since enumerated values can have a different type from entry to entry, eval may behave differently based on the type of a variable at runtime. For example, the filter foo < 5 will only work if “foo” can be interpreted at runtime as a number. It is not an error for “foo” to be a different type (in this example the boolean expression “foo < 5” would be false if foo is not a number).
The type() built-in function can be used at runtime to determine the type of a variable in a given entry.
eval can attempt to cast a variable to another type using the following cast operators.
int (equivalent to int64)
uint8
int8
uint16
int16
uint32
int32
uint64
int64
float
string
ip
mac
duration
time
type
bool
Additionally, eval will attempt to “promote” types that can be automatically cast. For example, if “foo” contains the string “56”, the expression foo < 3.14 will cause eval to attempt to promote foo to a floating-point number before the expression is evaluated.
Scope#
Eval has no variable scoping. All variables are “global”, regardless of if they are declared in a statement block or not.
Expressions#
Expressions are simply computations of values. The behavior of an expression depends on where in a program it is used.
Expressions follow the C-style form of expressions. For example, if (foo == "bar")... contains the boolean expression foo == "bar" which is used to compute a true/false value for determining if the body of the if statement should be executed or not.
The right-hand-side of an assignment is also an expression. For example, foo = type(bar); contains the expression type(bar), which is a built-in function that returns a value.
Expressions can be complex, and contain other expressions, just like in C-style programs. For example, ( a == b || b < 3.14 || ( type(b) == "string" && error == "none" )) is an expression made up of several, smaller expressions.
Note
When a program consists of only an expression, the program as a whole is treated as a filter. If the expression returns false (or returns the “zero” value of the type returned), the entry will be dropped.
Note
Expressions that operate on enumerated values that don’t exist have undefined behavior. If you aren’t sure that an EV will exist in each entry being processed, be sure to use the has() built-in to ensure the behavior you intend. For example:
if (has(myEV)) {
if (myEV == "foo") {
...
}
}
Type Promotion in Expressions#
Type promotion is used to satisfy boolean and arithmetic expressions of two or more enumerated values of different types. For example, the expression 1 + 2.14 contains an integer and a floating-point value. In this example, the integer will be “promoted” (converted) to a floating-point value, and the addition will result in a floating-point value.
Eval also supports implicit conversion of mixed types as a form of type polymorphism. This is also considered in the type promotion logic. For example, 5 + "10" can be seen to mean “the integer 5 plus the numeric representation of the string 10”. Eval will convert the string to the integer 10 in this example.
Not all types are compatible. Eval will attempt to determine a common real type for two given EVs using the following logic, in order:
If either is a Location, try to convert the other to a Location.
If either is an IP, try to convert the other to an IP.
If either is a MAC, try to convert the other to a MAC.
If either is a timestamp, try to convert the other to a timestamp.
If either is a float, try to convert the other to a float.
If either is an integer, try to convert the other to an integer.
If none of these cases apply, the EVs remain their original types.
Timestamps and durations have three special cases:
A timestamp plus a duration results in a timestamp.
A timestamp minus a duration results in a timestamp.
A timestamp minus a timestamp results in a duration.
When type promotion results in two integers, or two integers are provided in an expression, additional promotion logic is applied, conforming to a modified form of the ISO/IEC 9899:2011 C standard rank rules:
No two signed integer types shall have the same rank, even if they have the same representation.
The rank of a signed integer type shall be greater than the rank of any signed integer type with less precision.
An int is equivalent to an int64.
The rank of int64 shall be greater than the rank of int32, which shall be greater than the rank of int16, which shall be greater than the rank of int8.
The rank of any unsigned integer type shall equal the rank of the corresponding signed integer type, if any.
The rank of byte shall equal the rank of uint8.
The rank of bool shall be less than the rank of all other integer types.
For all integer types T1, T2, and T3, if T1 has greater rank than T2 and T2 has greater rank than T3, then T1 has greater rank than T3.
After integer promotion, the “usual arithmetic conversions” apply, in order:
If both operands have the same type, no further conversion is needed.
If both operands are of the same integer type (signed or unsigned), the operand with the type of lesser integer conversion rank is converted to the type of the operand with greater rank.
If the operand that has unsigned integer type has rank greater than or equal to the rank of the type of the other operand, the operand with signed integer type is converted to the type of the operand with unsigned integer type.
If the type of the operand with signed integer type can represent all of the values of the type of the operand with unsigned integer type, the operand with unsigned integer type is converted to the type of the operand with signed integer type.
Otherwise, both operands are converted to the unsigned integer type corresponding to the type of the operand with signed integer type.
Failed Type Promotions#
When type promotion fails, the expression is also considered to have “failed”. The result of an operation on a failed type promotion is undefined. This is important for conditional expressions. To avoid scenarios like this, the user is encouraged to use type() checks and casting before conditional comparison. For example, given the following eval program:
x = "foo";
y = 5;
if ( x < y ) {
output = "x is less than y";
}
The comparison result is undefined and the output = statement may be executed. To guard against this, add type/cast checks first:
x = "foo";
y = 5;
if ( type(int(x)) == "int" ) { // The cast succeeded, we can do the comparison
if ( x < y ) {
output = "x is less than y";
}
} else {
output = "x isn't a number";
}
Statements#
Statements control program execution. For example, if (foo == "bar") { ... } contains an “if” statement, which will determine how the program is to proceed.
Statements are separated by the { } punctuation in the case of “if” and “for”, and semicolons “;” for assignments. Currently, only “if”, “for”, and assignment statements are supported in eval.
Statements other than “if” and “for” must be ended with a semicolon ;.
if statements#
“if” statements specify conditional execution of code, according to a boolean expression. If the expression evaluates to “true”, the code will be executed. “if” statements can contain an “else” code path as well, for executing when the expression evaluates to “false”.
The boolean expression is always contained in parentheses ( ), and code blocks are contained in braces { }. For example:
if ( foo == "bar" ) {
output = "foo is bar!";
} else {
output = "foo is not bar!";
}
Eval also supports else if statements of arbitrary length. For example:
if ( foo == "bar" ) {
output = "foo is bar!";
} else if ( foo == "baz" ) {
output = "foo is baz!";
} else {
output = "foo is not bar!";
}
for statements#
“for” statements specify the repeated execution of a block. “for” statements use the C-style syntax of an initializer, condition, and post statement, and are contained in parentheses ( ) and separated by semicolons ;. Code blocks are contained in braces { }. For example, to iterate 10 times over a code block:
for ( i = 0; i < 10; i++ ) {
count = i;
}
break statements#
A “break” statement stops the execution of the innermost “for” statement it is contained in. “break” can only be used inside a “for” statement code block. Execution continues at the statement immediately after the “for” statement. For example:
/* for loop up to 10, but break after 5 iterations */
for ( i = 0; i < 10; i++ ) {
if ( i == 5 ) {
break;
}
}
continue statements#
A “continue” statement begins the next iteration of the innermost “for” statement it is contained in. “continue” can only be used inside a “for” statement code block. Execution continues at the beginning of the next iteration of the “for” statement. For example:
/* for loop up to 10, but only count the last 8 */
for ( i = 0; i < 10; i++ ) {
if ( i < 2 ) {
continue;
}
count++;
}
return statements#
The return keyword ends execution of the eval program and determines if the entry being executed on is dropped or passed in the query pipeline. return takes a single expression. Returning true will pass the entry, false, will drop it. For example:
if ( foo == "bar" && bar == "baz" ) {
return true;
}
return false;
Built-in functions#
Convenience#
delete#
function delete(key string)
Deletes the given enumerated value.
has#
function has(ev string) bool
Returns true if the given EV exists.
in#
function in(ev string, string...) bool
Returns true if any of the given strings match the given EV.
Example
vegetable = "potato";
hasPotato = in(vegetable, "turnip", "potato", "cabbage");
unix#
function unix(ev) int
Returns a UNIX time of the given enumerated value.
len#
function len(<expression>) int
Return the length of the given expression or enumerated value.
log#
function log(<expression>)
Log the given expression to the “gravwell” tag.
match#
function match(input string, pattern string) bool
Returns true if the regular expression in pattern matches the given input. Regular expressions use the RE2 syntax.
printf#
function printf(format string, <expression>...) string
Formats a string according to the given format and any number of expressions. printf uses the formatting rules and verbs defined in Golang’s fmt.Printf() function.
set_data#
function set_data(data <expression>)
Overwrites the entry’s DATA field. This is different than creating an enumerated value named DATA.
String manipulation#
strings_count#
function strings_count(input string, pattern string) int
Counts the number of non-overlapping instances of pattern in the given input. If pattern is an empty string, strings_count returns 1 + the number of Unicode code points in the input.
strings_hasprefix#
function strings_hasprefix(input string, prefix string) bool
Returns true if the string input begins with prefix.
strings_hassuffix#
function strings_hassuffix(input string, suffix string) bool
Returns true if the string input ends with suffix.
strings_index#
function strings_index(input string, pattern string) int
Returns the index of the first instance of pattern in the input, or -1 if pattern is not in input.
strings_replace#
function strings_replace(input string, old string, new string, n int) string
Replace returns a copy of the string s with the first n non-overlapping instances of old replaced by new. If n is -1, there is no limit on the number of replacements.
strings_tolower#
function strings_tolower(input string) string
Returns the input with all letters mapped to their lower case.
strings_totitle#
function strings_totitle(input string) string
Returns the input with all letters mapped to their Unicode title case.
strings_toupper#
function strings_toupper(input string) string
Returns the input with all letters mapped to their upper case.
strings_trim#
function strings_trim(input string, cutset string) string
Returns the input with all leading and trailing code points in cutset removed.
strings_trimleft#
function strings_trimleft(input string, cutset string) string
Returns the input with all leading code points in the cutset removed.
strings_trimprefix#
function strings_trimprefix(input string, prefix string) string
Returns the input with the leading prefix removed.
strings_trimright#
function strings_trimright(input string, cutset string) string
Returns the input with all trailing code points in the cutset removed.
strings_trimspace#
function strings_trimspace(input string) string
Returns the input with all leading and trailing whitespace removed.
strings_trimsuffix#
function strings_trimsuffix(input string, suffix string) string
Returns the input with the trailing suffix removed.
pretty_size#
function pretty_size(input string) string
Converts a number to an abreviated pretty printed size, 1234567 becomes “1.18 MB”.
pretty_count#
function pretty_count(input string) string
Converts a number to an abreviated pretty printed magnitude, 1234567 becomes “1.24 M”.
pretty_rate#
function pretty_rate(number, duration) string
Converts a number to an abreviated pretty printed rate in bytes, kilobytes, or megabytes per second given a magnitude and duration; “pretty_rate(1234567, “2s”)” becomes “588.87 KB/s”.
pretty_line_rate#
function pretty_line_rate(number, duration) string
Converts a number to an abreviated pretty printed line rate in bits, kilobits, and megabits per second given a magnitude and duration; “pretty_line_rate(1234567, “2s”)” becomes “4.71 Mb/s”.
Hash#
hash_md5#
function hash_md5(input <expression>) string
Returns the MD5 sum of the given input.
hash_sha1#
function hash_sha1(input <expression>) string
Returns the SHA1 sum of the given input.
hash_sha256#
function hash_sha256(input <expression>) string
Returns the SHA256 sum of the given input.
hash_sha512#
function hash_sha512(input <expression>) string
Returns the SHA512 sum of the given input.
Path#
path_base#
function path_base(path string) string
Returns the last elements of the given path. Trailing slashes are removed before extracting the last element.
path_clean#
function path_clean(path string) string
Returns the shortest path name equivalent to the given path. For example, “/../a/b/../././/c” would return as “/a/c”.
path_dir#
function path_dir(path string) string
Returns all but the last element of path.
path_ext#
function path_ext(path string) string
Returns the file name extension used by path.
path_isabs#
function path_isabs(path string) bool
Returns true if the path is absolute (no relative path elements).
path_join#
function path_join(string...) string
Joins any number of path elements into a single path, adding slashes between elements.
path_match#
function path_match(pattern string, name string) bool
Returns true if the name matches the given pattern.
The pattern syntax is identical to Golang’s path.Match() function.
Random#
rand_float#
function rand_float() float
Returns a random 64-bit float in the interval [0.0,1.0)
rand_int#
function rand_int() int
Returns a randomm, non-negative, 64-bit integer.
rand_intn#
function rand_intn(n int) int
Returns a random, non-negative, 64-bit integer in the range [0,n).
rand_normal#
function rand_normal() float
Returns a normally distributed 64-bit float with standard normal distribution (mean = 0, stddev = 1). To produce a different normal distribution, you can adjust the output using:
sample = rand_normal() * desiredStdDev + desiredMean;
Encoding#
base64_decode#
function base64_decode(input string) string
Returns the decoded form of the given base64 input.
base64_encode#
function base64_encode(input string) string
Returns the encoded base64 form of the given input.
hex_decode#
function hex_decode(input string) string
Returns the decoded form of the given hexadecimal string.
hex_encode#
function hex_encode(input string) string
Returns the encoded hexadecimal form of the given input.
html_escape#
function html_escape(input string) string
Returns the input with special characters like “<” changed to their escaped HTML form. It escapes only five characters: <, >, &, ‘, and “.
html_unescape#
function html_unescape(input string) string
Returns the input with escaped HTML entities such as “<” in their unescaped form.
json#
function json(input string) bool
Returns true if the given input is valid JSON.
json_append#
function json_append(array string, value <expression>) string
Appends the value to the JSON array given in input.
json_array#
function json_array(value <expression>) string
Returns a JSON array of the given value. The value’s type is evaluated at runtime and will map to the corresponding JSON type (object, array, bool, number, string), or a string if the type doesn’t map to a JSON type.
json_get#
function json_get(object string, key string) boo/float/string
Returns a typed item from the given object with the given key.
json_index#
function json_index(array string, index int) bool/float/string
Returns a typed item from the given array at the given index.
json_len#
function json_len(array string) int
Returns the length of the given JSON array.
json_object#
function json_object(key string, value <expression>) string
Returns a JSON object of the given key/value pair. The value’s type is evaluated at runtime and will map to the corresponding JSON type (object, array, bool, number, string), or a string if the type doesn’t map to a JSON type.
json_pretty#
function json_pretty(input string) string
Pretty prints the given JSON input.
json_set#
function json_set(object string, key string, value <expression>) string
Sets a key/value pair in the given object. The value’s type is evaluated at runtime and will map to the corresponding JSON type (object, array, bool, number, string), or a string if the type doesn’t map to a JSON type.
Math#
Note
Some math functions retain their legacy function names for backwards compatability.
ceil#
function ceil(x float) float
Returns the least integer value greater than or equal to x.
floor#
function floor(x float) float
Returns the greatest integer value less than or equal to x.
math_abs#
function math_abs(x float) float
Returns the absolut value of x.
math_ceil#
function math_ceil(x float) float
Same as ceil(). Returns the least integer value greater than or equal to x.
math_floor#
function math_floor(x float) float
Same as floor(). Returns the greatest integer value less than or equal to x.
math_log#
function math_log(x float) float
Returns the natural logarithm of x.
math_log10#
function math_log10(x float) float
Returns the decimal logarithm of x.
math_log2#
function math_log2(x float) float
Returns the binary logarithm of x.
math_max#
function math_max(x float, y float) float
Returns the larger of x or y.
math_min#
function math_min(x float, y float) float
Returns the smaller of x or y.
math_mod#
function math_mod(x float, y float) float
Returns the floating-point remainder of x/y.
math_pow#
function math_pow(x float, y float) float
Returns the base-x exponential of y.
math_remainder#
function math_remainder(x float, y float) float
Returns the IEEE 754 floating-point remainder of x/y.
math_round#
function math_round(x float) float
Same as round(). Returns the nearest integer, rounding half away from zero.
math_roundtoeven#
function math_roundtoeven(x float) float
Returns the nearest integer, rounding ties to even.
math_sqrt#
function math_sqrt(x float) float
Returns the square root of x.
math_trunc#
function math_trunc(x float) float
Returns the integer value of x by dropping decimal digits. For example, 3.1 and 3.9 will both return 3.
round#
function round(x float) float
Returns the nearest integer, rounding half away from zero.
Acceleration and eval#
Eval is capable of producing acceleration hints for the fulltext acceleration engine when used as a filter. For example:
tag=gravwell syslog Appname | eval ( Appname == "webserver" || Appname == "indexer" )
This query will provide the acceleration hints of “webserver” OR “indexer” to the fulltext engine. By using eval, complex acceleration hints can be composed that aren’t possible using extraction modules alone.
Syntax#
The eval syntax is expressed using a variant of Extended Backus-Naur Form (EBNF):
Program = ( "(" Expression ")" EOF ) | ( "(" Vars StatementList ")" EOF ) | ( "(" StatementList ")" EOF ) | ( "(" Assignment ")" EOF ) | ( Expression EOF ) | ( Vars StatementList EOF ) | ( StatementList EOF ) | ( Assignment EOF )
Vars = VarSpec { VarSpec }
VarSpec = ( "var" VarSpecAssignment { "," VarSpecAssignment } ";" ) | ( "map" AssignmentIdentifier ";" )
VarSpecAssignment = AssignmentIdentifier [ "=" Expression ]
StatementList = Statement { Statement }
Statement = ( "if" "(" Expression ")" Statement "else" Statement ) | ( "if" "(" Expression ")" Statement ) | ( "for" "(" Assignment ";" Expression ";" Assignment ")" "{" StatementList "}" ) | "{" StatementList "}" | Function ";" | Assignment ";" | "return" Expression ";" | "break" ";" | "continue" ";" | ";"
Assignment = ( AssignmentIdentifier "[" Expression "]" "=" Expression ) | ( AssignmentIdentifier "=" Expression ) | Expression
Expression = ( LogicalOrExpression "?" Expression ":" LogicalOrExpression ) | LogicalOrExpression
LogicalOrExpression = LogicalAndExpression { LogicalOrOp LogicalAndExpression }
LogicalAndExpression = InclusiveOrExpression { LogicalAndOp InclusiveOrExpression }
InclusiveOrExpression = ExclusiveOrExpression { BitwiseOrOp ExclusiveOrExpression }
ExclusiveOrExpression = AndExpression { ExclusiveOrOp AndExpression }
AndExpression = EqualityExpression { BitwiseAndOp EqualityExpression }
EqualityExpression = RelationalExpression { EqualityOp RelationalExpression }
RelationalExpression = ShiftExpression { RelationalOp ShiftExpression }
ShiftExpression = AdditiveExpression { ShiftOp AdditiveExpression }
AdditiveExpression = MultiplicativeExpression { AdditiveOp MultiplicativeExpression }
MultiplicativeExpression = UnaryExpression { MultiplicativeOp UnaryExpression }
UnaryExpression = UnaryOp PostfixExpression | PostfixExpression
PostfixExpression = PrimaryExpression [ PostfixOp ]
PrimaryExpression = NestedExpression | ( Identifier "[" Expression "]" ) | Identifier | Literal
NestedExpression = ( Function ) | ( Cast "(" Expression ")" ) | ( "(" Expression ")" )
Literal = DecimalLiteral | FloatLiteral | StringLiteral | "true" | "false"
Function = FunctionName "(" [ Expression { "," Expression } ] ")"
LogicalOrOp = "||"
LogicalAndOp = "&&"
ExclusiveOrOp = "^"
BitwiseOrOp = "|"
BitwiseAndOp = "&"
UnaryOp = "-" | "!"
PostfixOp = "++" | "--"
EqualityOp = "==" | "!=" | "~" | "!~"
RelationalOp = "<" | ">" | "<=" | ">="
ShiftOp = "<<" | ">>"
AdditiveOp = "+" | "-"
MultiplicativeOp = "*" | "/" | "%"
Cast = "int" | "float" | "string" | "mac" | "ip" | "time" | "duration" | "type" | "bool" | "location" | "byte" | "int8" | "uint8" | "int16" | "uint16" | "int32" | "uint32" | "int64" | "uint64"
Legacy Eval#
There is a legacy version of eval that you may still see in older queries. For more details, see the Legacy eval page for reference.