JavaScript expressions

Many Mapshaper commands take a JS expression as an argument or option. Expressions let you read and write per-feature attributes, derive new fields, filter records, sort, generate templated commands, and inspect layer-level metadata. The same expression syntax and execution context are reused across commands, so once you've learned the shape of a -each expression you can use it almost everywhere.

mapshaper counties.shp \
  -each 'STATE_FIPS = COUNTY_FIPS.substr(0, 2),
         AREA_KM2 = this.area / 1e6' \
  -o out.shp

Expressions are plain JavaScript. They can use any built-in language feature (arithmetic, string methods, conditionals, regex, etc.). Some commands also expect the expression to return a particular kind of value — -filter and -inspect expect true or false, -sort expects a sort key, -split expects a group identifier, and so on.

Where expressions appear

These five flavors — feature, calc, pair, layer and template — share most of their context but differ in which variables are available and which functions are in scope.

The execution context

Inside any feature-level expression you have access to:

• Field names as bare variables. Reading a field name returns its value. Assigning to a field name updates the current feature's record (and creates the field on first use). If a field name is not a valid JavaScript identifier (e.g. it contains spaces or starts with a digit), use d["field name"] to reference it.
• this, the feature proxy. Provides geometry-derived properties (this.area, this.bbox, etc.) and read/write access to the feature's properties, geojson and coordinates.
• d, a reference to the data record (the same object as this.properties).
• global, an object that persists across commands. Variables created by -define, by assignment in a -calc expression, or by writing to global.foo = ... inside -each end up here. Values set by -vars and -defaults live in a separate templating scope (read by {{X}} substitution) and are not visible by bare name in expressions; use -define if you want a value reachable from both {{X}} and JS expressions.
• console.log() for printing values to stderr while debugging.
• Built-in helpers (see Helper functions below).
• User helpers loaded by -define, -include or -require.

If a name is referenced but not present in any of the above, JavaScript treats it as undefined, not an error. This is convenient when chaining expressions across heterogeneous datasets but can mask typos — double-check field names with mapshaper -info if a -filter returns a suspiciously empty result.

Field assignment

Assigning to a bare name creates or updates a data field on the current feature:

mapshaper counties.shp -each 'POP_DENSITY = POPULATION / (this.area / 1e6)' -o

Bare assignments like POP_DENSITY = ... will create the data table if the layer doesn't already have one. Assignments routed through this.properties.X = ... or d.X = ... only update an existing data table — prefer a bare assignment if you're not sure the layer has one yet.

To delete a field, use the JS delete operator:

mapshaper states.shp -each 'delete STATE_NAME, delete GEOID' -o

To replace the entire record, assign to this.properties:

mapshaper states.shp -each 'this.properties = {FID: this.id, NAME: NAME}' -o

Multiple statements

Use commas to evaluate multiple sub-expressions. The value of the whole expression is the value of the last sub-expression (relevant for -filter, -sort, -split):

mapshaper data.csv -each 'A = parseInt(A), B = A * 2, C = A + B'

Inside command files, you can also break a long expression across lines with \:

-each '
  STATE_FIPS = COUNTY_FIPS.substr(0, 2),  \
  AREA_KM2   = this.area / 1e6,           \
  CENTROID_X = this.centroidX,            \
  CENTROID_Y = this.centroidY
'

Feature properties (this)

this is a proxy for the current feature. It gives you geometry-derived properties and a few editing affordances. The properties below are read-only unless the description says otherwise.

All layer types

Polygon, polyline and point layers (with geometry)

Polygon-only

Polyline-only

Point-only

Why it matters for unprojected data: this.area and this.length use spherical (not planar or ellipsoidal) geometry on lat/long datasets. Results are in square meters / meters and accurate to within ~0.5% for most use cases. If you need ellipsoidal accuracy, project first with -proj.

Layer-level properties (this.layer)

this.layer exposes information about the layer the feature belongs to. Useful in expressions that need to know about other features:

Helper functions

These are always in scope inside feature expressions:

• round(num [, decimals]) — Round to N decimal places (default 0). Faster and easier than Math.round.
• sprintf(fmt, ...) — printf-style formatter (uses printj syntax).
• format_dms(coord [, fmt]) — Format a number as a degrees/minutes/seconds string. Common formats: 'DD° MM′ SS.SSSSS″ [NS]', 'DdMmSs [EW]', '[+-]DDDMM.MMMMM', '[-]DD.DDDDD°'.
• parse_dms(string [, fmt]) — Parse a DMS string back to a number.
• blend(c1, c2, ...) — Mix CSS color strings together (returns a hex string).
• console.log(...) — Write to stderr.

JavaScript's built-in Math, JSON, Number, String, Array, Date, Object etc. are all available. Node-specific globals like process, require and setTimeout are not.

Calc expressions

-calc and any command's calc= option use the same context as -each plus a set of aggregate functions that operate over the entire group of features (or the entire layer for -calc). Each aggregate function takes a per-feature expression and reduces it to a single value across the group.

Argument expressions use the same syntax as -each, so per-feature properties and helpers are available:

mapshaper counties.shp \
  -calc 'TOTAL_POP = sum(POP),
         MEAN_AREA_KM2 = sum(this.area / 1e6) / count(),
         TOP_DENSITY = max(POP / this.area)'

Calc expressions can also use assignments to expose values to subsequent commands via the global namespace (see Sharing state across commands below).

Pair expressions (A and B)

The -lines where= and each= options operate on path segments shared between two adjacent features. Inside these expressions:

• A is the feature on one side of the path
• B is the feature on the other side, or null for outer boundaries

Both A and B give you the full set of feature properties (A.properties, A.area, A.id, etc.).

# Keep only inner boundaries between two different states
mapshaper counties.shp \
  -lines where='B && A.STATE != B.STATE' \
  -o state-borders.shp

Layer-level expressions (-if, -define)

The -if family and -define evaluate against the current command's target layer(s), not per feature. The context exposes:

• target — the proxy for the single target layer (only set when there's exactly one target)
• targets — an array-like of layer proxies, also indexable by name (targets.states)
• layer_name, data, type, size, empty, bbox
• field_exists(name), field_type(name), field_includes(name, value)
• layer_exists(name [, geometry_type])
• file_exists(path)
• global — the shared variable namespace

mapshaper data.csv \
  -calc 'N = count()' \
  -if 'global.N < 5' -print 'LOW SAMPLE SIZE, STOPPING' -stop -endif

Each entry in targets exposes useful summary stats from -info: layer_name, feature_count, null_shape_count, null_data_count, bbox, proj4. Reading targets[0].geojson returns the layer as a GeoJSON FeatureCollection; assigning to it replaces the layer with the FeatureCollection you provide.

Template expressions (-run)

-run accepts either a path to a command file or a string containing one or more curly-brace template expressions. Each {...} is evaluated as a JS expression and substituted into the resulting command string before Mapshaper parses it.

# Project to a transverse Mercator centred on the layer
mapshaper -i country.shp -require projection.js \
  -run '-proj {tmerc(target.bbox)}' -o

Inside the curly braces you have:

• target and targets (same as -if)
• io.ifile(filename, data) — spill data to a temp file and yield its path, useful for piping computed JSON back into -i
• Anything loaded by -require or -define

Bare function calls outside curly braces are also evaluated directly, so -run 'tmerc(target.bbox)' works the same as -run '{tmerc(target.bbox)}' when the function name was loaded via -require.

Loading helpers

Three commands extend the expression context with your own variables and helpers:

• -define takes an inline JS expression and stores any assignments on the global namespace. Good for one-liners.
• -include loads a .js file containing a single object literal; each property of that object becomes a variable in subsequent expressions.
• -require loads an installed npm module or a local module file. With alias=foo the module is bound to that name; without an alias, the module's exported names are added directly to the context.

mapshaper data.json \
  -require ./helpers.mjs \
  -each 'displayname = formatName(d)' \
  -o data.json

mapshaper -define 'KM_PER_MILE = 1.609344' \
  routes.geojson \
  -each 'KM = MILES * global.KM_PER_MILE' \
  -o

Sharing state across commands

Mapshaper has two scopes for values that persist between commands. They share a name lookup for {{X}} substitution but are otherwise independent.

• Expression scope (global) — written by -define, -include, -require, -colorizer, and -calc assignments (e.g. N = count()) or any global.foo = ... inside -each. Values can be any JavaScript value (numbers, strings, functions, objects). Read by JS expressions as bare names, and as global.X everywhere. {{X}} substitution falls back to this scope, so -define base = "out" → -o {{base}}.geojson and -calc 'N = count()' → -if '{{N}} > 100' work as you'd expect.
• Templating scope — written by -vars and -defaults. Values must be primitives (string / number / boolean / null) and are validated at write time. Read by {{X}} substitution; {{X}} checks the templating scope first, then falls back to the expression scope. Not visible by bare name in JS expressions — that's deliberate, so a string set by -vars N=5 can't silently coerce into arithmetic.

If you want one value usable in both contexts, set it once with -define. If you only need it in command strings, use -vars (or -defaults for command-file overridable defaults).

mapshaper counties.shp \
  -calc 'BIG = count("POP > 1000000")' \
  -if 'global.BIG > 0' \
    -filter 'POP > 1000000' \
    -o big-counties.shp \
  -endif

Common pitfalls

• Quoting. In bash/zsh, wrap expressions in single quotes so the shell doesn't expand !, $ or backticks. In Windows cmd.exe, use double quotes and escape inner quotes with backslashes. In PowerShell, prefer single quotes, or escape $ with a backtick.
• Type coercion from CSVs. Numeric-looking strings in CSVs are parsed as numbers by default; identifier-like strings (FIPS, ZIP) need string-fields= on -i to preserve leading zeros. See CSV practical notes.
• Field name collisions. A field called area, length, id etc. shadows the built-in property of the same name. Mapshaper prints a warning. Either rename the field with -rename-fields, or read the property via this.area rather than the bare name.
• Lat/long area surprises. this.area on an unprojected polygon returns square meters on a sphere, not square degrees. To get square kilometres, divide by 1e6. To get planar square degrees (e.g. for sanity checks), use this.planarArea.
• Centroids ignore holes. this.centroidX/Y is the centroid of the largest ring. For a labelling point that's guaranteed inside the polygon, use innerX/innerY.
• -each doesn't return values. Its expression is evaluated for side-effects only. Use -filter, -sort or -calc if you want the return value to drive behavior.
• Reserved names. this, d, _, global, console, target, targets and the helper function names listed above are not safe to use as field names.
• Auto-vivification of fields. Assigning to a name that isn't a known field creates a new field on every record. If you only want to set a field on some records, wrap it in a conditional and assign explicit null for the others, otherwise downstream readers may see undefined instead of a real null.

Examples

# Add two derived fields
mapshaper counties.shp \
  -each 'STATE_FIPS = COUNTY_FIPS.substr(0, 2),
         AREA_KM2 = round(this.area / 1e6, 2)' \
  -o out.shp

# Drop features outside a date window
mapshaper events.csv \
  -filter 'new Date(DATE) >= new Date("2020-01-01")' \
  -o recent.csv

# Sort polygons largest-first
mapshaper countries.geojson \
  -sort '-this.area' \
  -o sorted.geojson

# Look up one feature
mapshaper states.geojson -inspect 'NAME == "Delaware"'

# Aggregate stats during a dissolve
mapshaper counties.shp \
  -dissolve STATE calc='N = count(),
                        POP = sum(POP),
                        MEDIAN_INC = median(MEDIAN_INC)' \
  -o states.shp

# Conditional pipeline based on a calc result
mapshaper data.csv \
  -calc 'N = count()' \
  -if 'global.N == 0' -stop -endif \
  -o data.csv

# Filter shared boundaries
mapshaper counties.shp \
  -lines where='B && A.STATE != B.STATE' \
  -o state-borders.shp

# Per-feature styling: circle radius from POP, fill from an expression
mapshaper cities.geojson \
  -style r='Math.sqrt(POP) / 40' \
         fill='POP > 1e6 ? "#c33" : "#39c"' \
         opacity=0.7 \
  -o cities.svg

# Project to a layer-specific CRS
mapshaper -i country.shp -require ./projection.js \
  -run '-proj {tmerc(target.bbox)}' -o

See also

• -each — the canonical feature-expression command
• -filter
• -calc
• -define, -include, -require
• -run
• Basics — recipes that put expressions to work