Attributes reference

Every attribute in the tag DSL, with a one-line description and where it's covered.

The <compute> tags have attributes of their own (v, sep, as, width, fill, from, to, size, pattern, default). Each one belongs to a single tag and is covered where that tag is explained — see the compute reference.

Environment & config

Attribute	What it sets	See
version / v	The DSL version the file requires	Configuration
count	Number of records	Determinism
seed	RNG seed, for reproducibility	Determinism
local	Locale for template data — on <env> for the whole run, and on one <gen type="template"> to override it for that sequence alone	Template
inject	Custom interpolation marker	Output & formatting
mode	memory / disk — which engine family; stream is a legacy alias that forces engine 2; sequential is a fourth value that prev() requires	Large outputs, Expressions
engine	1 / 2 / 3 — force one engine (advanced)	Large outputs
comment	Free-form comment	Configuration

Sequences & dependencies

Attribute	What it sets	See
name	On <sequence>: its name. On <gen>: makes it a field, Sequence.Field. On <data> inside a sequence: a constant field, the only one that costs no draw	Sequences
parent	Parent filter, Parent.Value	Hierarchical dependencies
uniq	Combination that must be unique across all rows	Unique values
on / is	Subject / branch key for <switch>	Switch
filter	On <gen type="pool">: which members this row may draw from	Coherent records
of	On <gen type="running">: the column to accumulate. On <gen type="stat">: the column to summarise. On <gen type="date">: the column to measure from	Running total, Statistic
plus	On <gen type="date" of="…">: how far from that column — 7d, 3..10d, 1..3mo, -10..-3d; a bare number means days	An interval
op	On <gen type="stat">: which statistic — sum, mean, median, min, max, count or stddev	Statistic
expr	On <gen type="formula">: the arithmetic this column is, written the way an if= condition is written	Formula
that	On <assert>: the condition the finished run must satisfy, in the if= language	Self-checking configs
says	On <assert>: the sentence a reader is given when it does not hold	Self-checking configs
reset	On <gen type="running">: a column whose change restarts the total	Running total

Generator values

Attribute	What it sets	See
type	Which generator to use	Generators
value	The generator's main value (type-specific)	Generators
percent	Exact distribution of the values	Text
accumulate	Replace a repeat list with its running total, or say how a running column accumulates	Several values in a cell
alphabet	Named Unicode alphabet	Symbol
length	Output length or width	Number
first_zero	Allow a leading zero	Number
weekdays	Which weekdays a walked date axis keeps: mon..fri, sun,wed	Date
step	Counter stride, or how far a walked date axis advances: 15m, 1h30m, 3mo	Counters, Date
regex_max_length	Length cap for a regex	Regex
include / exclude	Keep or drop values from the pool	Number
decimals	Digits after the decimal point	Number
oldest / youngest	Birthday age window	Date
format	Date output format	Date
from / to	Range endpoints, given separately	Date
precision	Step size for a date-time range	Date
range	Date range for date.range	Template

Statistical shape

Attribute	What it sets	See
distribution	Named distribution (normal, zipf, …)	Distributions
min / max	Clip the drawn values to a range	Distributions
missing	Share of rows left empty	Missing data
missing_as	How an empty cell is written	Missing data
missing_when	Which rows may go missing (MAR / MNAR)	Missing data
anomaly	Share of rows turned into outliers	Anomalies
anomaly_factor	How far an outlier is pushed	Anomalies
anomaly_flag	Answer column that marks the outliers	Anomalies

Each distribution takes its own parameters, and they are only read when distribution= names that one. Every distribution also accepts decimals, min and max. Each is explained, with a histogram, on the distributions guide.

distribution=	Parameters	What they mean
normal	mean sd	The centre and the spread
lognormal	meanlog sdlog	The centre and spread of the logarithm — the value itself is skewed right
exponential	rate	Events per unit of time; the mean is 1/rate
pareto	alpha xmin	Tail thickness, and the smallest possible value
weibull	shape scale	shape below 1 = early failures, above 1 = wear-out; scale sets the typical lifetime
poisson	lambda	Average count per interval (capped at 700)
zipf	n s	How many ranks, and how steeply they fall off
gamma	shape scale	Total wait for shape events that each take scale on average
beta	alpha beta	Pull toward 1 and toward 0 — the result is between 0 and 1

Timeseries

Attribute	What it sets	See
base	Starting level of the series	Timeseries
trend	Drift per step	Timeseries
period	Length of one seasonal cycle	Timeseries
amplitude	Height of the seasonal swing	Timeseries
peak_at	Which row the seasonal wave peaks on	Timeseries
noise	Random jitter added on top	Timeseries
noise_correlation	How much of one row's jitter carries into the next	Timeseries

Pattern (a drawing as the source)

Attribute	What it sets	See
points	Inline x,y pairs instead of a file	Pattern
upper / lower	Two boundary curves — a corridor	Pattern
mode	signal (a trajectory) / density (a distribution)	Pattern
y_range	min..max — the vertical scale (required)	Pattern
fit	low..high — where a drawing from src lands	Pattern
interp	linear / smooth / step between points	Pattern
spread	Widen the line into a band of ±N	Pattern
ink_threshold	How dark a PNG pixel has to be to count as ink	Pattern

mode is really three different readings that share a name. On <env> it does two jobs: memory / disk pick the engine family, while sequential is a promise about row ORDER — row N is computed after row N−1, which is what prev() needs and what pins the run to engine 1. On a pattern generator it picks what you're asking the drawing for.

The CLI's --mode covers only the engine-family half: --mode memory and --mode disk are the whole set, so sequential is reachable from the config and nowhere else.

Files & CSV

Attribute	What it sets	See
src	Path to a data file	File
column	CSV column (name or number)	File
header	Skip the first CSV row	File
delimiter	CSV separator	File
row	Linked-row key	File
weight	Frequency column for weighted rows	Coherent data
read	"quantile" — read the file as a sorted sample and land anywhere on it	File
sample	"exact" — sweep that distribution evenly instead of drawing from it	File

HTTP service

Attribute	What it sets	See
src	Service URL (the same attribute as the file path above)	HTTP service
in	Sequence whose value is sent with each row	HTTP service
on_error	fail (default) or empty when a request fails	HTTP service
timeout	Seconds to wait for a response (default 30)	HTTP service
secret	Key each request is signed with — env:, file: or a literal	HTTP service

Output & formatting

Attribute	What it sets	See
if	Display condition (an expression)	Output & formatting
pair	Paired marker for a literal </data>	Output & formatting
mask	Display mask (x/w/*)	Masks & case
case	Letter case (upper/lower/…)	Masks & case
order	Value order (random / sequential) — text, file and date only	Generators
cycle	With sequential: cycle or raise an error — the same three types	Generators
repeat / separator	Several values in one cell	Multiple values
lengths	Beside repeat="A..B": the share of rows that get each possible length, A first — an exact quota, not an approximation	Multiple values
distinct	No repeats inside one cell (needs repeat)	Multiple values
each	Repeat a line for each list element; on <assert>, a condition every row must satisfy	Relational tables, Self-checking configs
flag	Answer column that marks <mix> outliers	Mix