Uniform eXchange Format (UXF) is a plain text human readable optionally typed storage format. UXF may serve as a convenient alternative to csv, ini, json, sqlite, toml, xml, or yaml.

UXF is an open standard. The UXF software linked from this page is all free open source software.

The primary purpose of UXF is to make developers lives easier. It does this by providing a convenient, scalable, easy-to-use file format that can be used for most purposes, from configuration files to application data. And UXF-based formats are very easy to adapt to future requirements

• Datatypes
  • Table of Built-in Types
  • Terminology
  • Minimal empty UXF
  • Built-in Types
  • Custom Types
• Examples
• Libraries (Python)
• Imports
• BNF
• Supplementary
  • Vim Support
  • UXF Logo

UXF supports the following eleven built-in datatypes.

Note that it is also possible to represent Custom Types.

• A map key-value is collectively called an item.
• A “single” valued type (bool, bytes, date, datetime, int, str), is called a scalar.
• A “multi-” valued type (list, map, table) is called a collection.
• A list, map, or table which contains only scalar values is called a scalar list, scalar map, or scalar table, respectively.
• A ttype is the name of a table's user-defined type.

uxf 1.0
[]

Every UXF file consists of a single header line (starting uxf 1.0, optionally followed by custom text). This may be followed by an optional file-level comment, then any ttype (table type) imports, then any ttype definitions. After this comes the data in the form of a single list, map, or table in which all the values are stored. The data must be present even if it is merely an empty list (as here), an empty map (e.g., {}), or an empty table. Since lists, maps, and tables can be nested inside each other, the UXF format is extremely flexible.

Map keys (i.e., ktype) may only be of types bytes, date, datetime, int, and str.

List, map, and table values may be of any type (including nested maps, lists, and tables), unless constrained to a specific type. If constrained to a specific vtype, the vtype may be any built-in type (as listed above, except null), or any user-defined ttype, and the corresponding value or values must be any valid value for the specified type, or ? (null).

Lists and tables preserve the order in which values are appended. So the first value is at index/row 0, the second at index/row 1, etc. Maps are unordered.

A table starts with a ttype. Next comes the table's values. The number of values in any given row is equal to the number of field names in the ttype.

Lists, maps, tables, and ttype definitions may begin with a comment. And lists, maps, and tables may optionally by typed as indicated above. (See also the examples below and the BNF near the end).

Strings may not include &, < or >, so if they are needed, they must be replaced by the XML/HTML escapes &, <, and > respectively.

Where whitespace is allowed (or required) it may consist of one or more spaces, tabs, or newlines in any combination.

If you don't want to be committed to a particular UXF type, just use a str and do whatever conversion you want, or use a Custom Type.

There are two common approaches to handling custom types in UXF. Both allow for UXFs to remain round-trip readable and writeable even by UXF processors that aren't aware of the use of custom types as such.

Here, we'll look at both approaches for two different custom types, a point and an enumeration.

uxf 1.0
[
  {<Point> [1.4 9.8]} {<Point> [-0.7 3.0]} {<Point> [2.1 -6.3]}
  <TrafficLightGreen> <TrafficLightAmber> <TrafficLightRed>
]

This first approach shows three points, each represented by a map with a str indicating the custom type (“Point”), and using lists of two reals for the x and y coordinates. The example also shows a traffic light enumeration each represented by a str.

uxf 1.0
[
  {<Point> [1.4 9.8 -0.7 3.0 2.1 -6.3]}
  <TrafficLightGreen> <TrafficLightAmber> <TrafficLightRed>
]

Since we have multiple points we've changed to a single map with a list of point values. This is more compact but assumes that the reading application knows that points come in pairs.

A UXF processor has no knowledge of these representations of points or enumerations, but will handle both seamlessly since they are both represented in terms of built-in UXF types. Nonetheless, an application that reads such UXF data can recognize and convert to and from these representations to and from the actual types.

uxf 1.0
=Point x:real y:real
=TrafficLightGreen
=TrafficLightAmber
=TrafficLightRed
[
  (Point 1.4 9.8 -0.7 3.0 2.1 -6.3)
  (TrafficLightGreen) (TrafficLightAmber) (TrafficLightRed)
]

This second approach uses four ttypes (custom table types). For the Point we specify it as having two real fields (so the processor now knows that Points have two real values). And for the enumeration we used three separate fieldless tables.

Using tables has the advantage that we can represent any number of values of a particular ttype in a single table (including just one, or even none), thus cutting down on repetitive text. Here, the Point table has three Points (rows). And some UXF processor libraries will be able to return table values as custom types. (For example, the Python UXF library would return these as custom class instances—as “editable tuples”.)

If many applications need to use the same ttypes, it may make sense to create some shared ttype definitions. See Imports for how to do this.

uxf 1.0
{}

We saw earlier an example of a minimal UXF file with an empty list; here we have one with an empty map.

uxf 1.0
=Pair first second
(Pair)

Here is a UXF with a ttype specifying a Pair that has two fields each of which can hold any UXF value (including nested collections). In this case the data is a single empty Pair table.

uxf 1.0
=Pair first second
(Pair (Pair 1 2) (Pair 3 (Pair 4 5)))

And here is a UXF with a single Pair table that contains two nested Pair tables, the second of which itself contains a nested pair.

Although widely used, the CSV format is not standardized and has a number of problems. UXF is a standardized alternative that can distinguish fieldnames from data rows, can handle multiline text (including text with commas and quotes) without formality, and can store one—or more—tables in a single UXF file.

Date,Price,Quantity,ID,Description
"2022-09-21",3.99,2,"CH1-A2","Chisels (pair), 1in & 1¼in"
"2022-10-02",4.49,1,"HV2-K9","Hammer, 2lb"
"2022-10-02",5.89,1,"SX4-D1","Eversure Sealant, 13-floz"

The most obvious translation would be to a list of lists:

uxf 1.0
[
  [<Price List> <Date> <Price> <Quantity> <ID> <Description>]
  [2022-09-21 3.99 2 <CH1-A2> <Chisels (pair), 1in &amp; 1¼in>]
  [2022-10-02 4.49 1 <HV2-K9> <Hammer, 2lb>]
  [2022-10-02 5.89 1 <SX4-D1> <Eversure Sealant, 13-floz>]
]

This is perfectly valid. However, it has the same problem as .csv files: is the first row data values or column titles? (For software this isn't always obvious, for example, if all the values are strings.) Not to mention the fact that we have to use a nested list of lists. Nonetheless it is an improvement, since unlike the .csv representation, every value has a concrete type (all strs for the first row, and date, real, int, str, str, for the subsequent rows).

The most appropriate UXF equivalent is to use a UXF table:

uxf 1.0
=PriceList Date Price Quantity ID Description
(PriceList
  2022-09-21 3.99 2 <CH1-A2> <Chisels (pair), 1in &amp; 1¼in> 
  2022-10-02 4.49 1 <HV2-K9> <Hammer, 2lb> 
  2022-10-02 5.89 1 <SX4-D1> <Eversure Sealant, 13-floz> 
)

When one or more tables are used each one's ttype (table type) must be defined at the start of the .uxf file. A ttype definition begins with an = sign followed by the ttype (i.e., the table name), followed by one or more fields. A field consists of a name optionally followed by a : and then a type (here only names are given).

Both table and field names are user chosen and consist of 1-60 letters, digits, or underscores, starting with a letter or underscore. No table or field name may be the same as any built-in type name, so no table or field can be called bool, bytes, date, datetime, int, list, map, null, real, str, or table. (But Date, DateTime, and Real or real_ are fine, since names are case-sensitive and none of the built-in types contains an underscore or uses uppercase letters.) If whitespace is wanted one convention is to use underscores in their place.

Once we have defined a ttype we can use it.

Here, we've created a single table whose ttype is “PriceList”. There's no need to group rows into lines as we've done here (although doing so is common and easier for human readability), since the UXF processor knows how many values go into each row based on the number of field names. In this example, the UXF processor will treat every five values as a single record (row) since the ttype has five fields.

This is already an improvement on .csv—we know the table's name and field names, and could easily store two or more tables (as we'll see later). Although the UXF processor will correctly determine the field types, what if we want to constrain each field's value to a particular type?

uxf 1.0 Price List
=PriceList Date:date Price:real Quantity:int ID:str Description:str
(PriceList
  2022-09-21 3.99 2 <CH1-A2> <Chisels (pair), 1in &amp; 1¼in> 
  2022-10-02 4.49 1 <HV2-K9> <Hammer, 2lb> 
  2022-10-02 5.89 1 <SX4-D1> <Eversure Sealant, 13-floz> 
)

Here we've added a custom file description in the header, and we've also added field types to the ttype definition. When types are specified, the UXF processor is expected to be able to check that each value is of the correct type. Omit the type altogether (as in the earliler examples) to indicate any valid table type.

uxf 1.0 Price List
=PriceList Date:date Price:real Quantity:int ID:str Description:str
(PriceList)

Just for completeness, here's an example of an empty price list table.

Windows .ini format (and Unix's often similar .conf format) are commonly used but unstandardized formats. UXF can be used as a more reliable and easier to use alternative.

shapename = Hexagon
zoom = 150
showtoolbar = False
[Window1]
x=615
y=252
width=592
height=636
scale=1.1
[Window2]
x=28
y=42
width=140
height=81
scale=1.0
[Window3]
x=57
y=98
width=89
height=22
scale=0.5
[Files]
current=test1.uxf
recent1=/tmp/test2.uxf
recent2=C:\Users\mark\test3.uxf

This first equivalent is a simplistic conversion that we'll improve in stages.

uxf 1.0 MyApp 1.2.0 Config
=Files Kind Filename
{
  <General> {
    <shapename> <Hexagon>
    <zoom> 150
    <showtoolbar> no
  }
  <Window1> {
    <x> 615
    <y> 252
    <width> 592
    <height> 636
    <scale> 1.1
  }
  <Window2> {
    <x> 28
    <y> 42
    <width> 140
    <height> 81
    <scale> 1.0
  }
  <Window3> {
    <x> 57
    <y> 98
    <width> 89
    <height> 22
    <scale> 0.5
  }
  <Files> (Files
    <current> <test1.uxf> 
    <recent1> </tmp/test2.uxf> 
    <recent2> <C:\Users\mark\test3.uxf> 
  )
}

UXF accepts no and yes for bool false and true repectively. (0 and 1 cannot be used as bools since the UXF processor would interpret them as ints.)

For configuration data it is often convenient to use maps with name keys and data values. In this case the overall data is a map which contains each configuration section. The values of each of the first two of the map's keys are themselves maps. But for the third key's value we use a table. The table's ttype is defined at the start and consists of two untyped fields.

Of course, we can nest as deep as we like and mix lists and maps. For example, here's an alternative:

uxf 1.0 MyApp 1.2.0 Config
=pos x:int y:int
=size width:int height:int
{
  <General> {#<Miscellaneous settings>
    <shapename> <Hexagon> <zoom> 150 <showtoolbar> no <Files> {
      <current> <test1.uxf>
      <recent> [#<From most to least recent>
      </tmp/test2.uxf> <C:\Users\mark\test3.uxf>]
    }
  }
  <Window1> {#<Window dimensions and scales> str
    <pos> (pos 615 252)
    <size> (size 592 636)
    <scale> 1.1
  }
  <Window2> {
    <pos> (pos 28 42)
    <size> (size 140 81)
    <scale> 1.0
  }
  <Window3> {
    <pos> (pos 57 98)
    <size> (size 89 22)
    <scale> 0.5
  }
}

Here, we've laid out the General and Window maps more compactly. We've also moved the Files into General and changed the Files from a table to a two-item map with the second item's value being a list of filenames. We've also changed the x, y coordinates and the width and height into “pos” and “size” tables. Notice that for each of these tables we've defined their ttype to include both field names and types.

We've also added some example comments to two of the maps. A comment is a # immediately followed by a str. In the data, a comment may only be placed at the start of a list before the optional vtype or the first value, or at the start of a map before the optional ktype or the first key, or at the start of a table before the ttype name.

uxf 1.0 MyApp 1.2.0 Config
=pos x:int y:int
=size width:int height:int
{#<We want str keys and map values> str map
  <General> {#<We want str keys and any values> str
    <shapename> <Hexagon> <zoom> 150 <showtoolbar> no <Files> {str
      <current> <test1.uxf>
      <recent> [#<From most to least recent> str
      </tmp/test2.uxf> <C:\Users\mark\test3.uxf>]
    }
  }
  <Windows> {#<Window dimensions and scales>
    <pos> (pos 615 252 28 42 57 98)
    <size> (size 592 636 140 81 89 22)
    <scale> [1.1 1.0 0.5]
  }
}

Here we've added some types. The outermost map must have str keys and map values, and the General, Files, and Window maps must all have str keys and any values. For maps we may specify the key and value types, or just the key type, or neither. We've also specified that the recent files list's values must be strs.

Notice that instead of individual “Windows” entries we've just used one. Since “pos” and “size” are tables they can have as many rows as we like, in this case three (since each row has two fields based on each table's ttype).

uxf 1.0 MyApp 1.2.0 Config
=#<Window dimensions> Geometry x:int y:int width:int height:int scale:real
{#<Notes on this configuration file format> str map
  <General> {#<Miscellaneous settings> str
    <shapename> <Hexagon> <zoom> 150 <showtoolbar> no <Files> {str
      <current> <test1.uxf>
      <recent> [#<From most to least recent> str
      </tmp/test2.uxf> <C:\Users\mark\test3.uxf>]
    }
  }
  <Windows> (#<Window dimensions and scales> Geometry
     615 252 592 636 1.1
     28 42 140 81 1.0
     57 98 89 22 0.5
  )
}

For this variation we've gathered all the window data into a single table type and laid it out for human readability. We could, of course, have just written it as (Geometry 615 252 592 636 1.1 28 42 140 81 1.0 57 98 89 22 0.5).

Database files aren't normally human readable and usually require specialized tools to read and modify their contents. Yet many databases are relatively small (both in size and number of tables), and would be more convenient to work with if human readable. For these, UXF format provides a viable alternative.

A UXF equivalent to a database of tables can easily be created using a list of tables:

uxf 1.0 MyApp Data
=Customers CID Company Address Contact Email
=Invoices INUM CID Raised_Date Due_Date Paid Description
=Items IID INUM Delivery_Date Unit_Price Quantity Description
[#<There is a 1:M relationship between the Invoices and Items tables>
  (Customers
    50 <Best People> <123 Somewhere> <John Doe> <[email protected]> 
    19 <Supersuppliers> ? <Jane Doe> <[email protected]> 
  )
  (Invoices
    152 50 2022-01-17 2022-02-17 no <COD> 
    153 19 2022-01-19 2022-02-19 yes <> 
  )
  (Items
    1839 152 2022-01-16 29.99 2 <Bales of hay> 
    1840 152 2022-01-16 5.98 3 <Straps> 
    1620 153 2022-01-19 11.5 1 <Washers (1-in)> 
  )
]

Here we have a list of tables representing three database tables. The list begins with a comment.

Notice that the second customer has a null (?) address and the second invoice has an empty description.

uxf 1.0 MyApp Data
#<It is also possible to have one overall comment at the beginning,
after the uxf header and before any ttype definitions or the data.>
=Customers CID:int Company:str Address:str Contact:str Email:str
=Invoices INUM:int CID:int Raised_Date:date Due_Date:date Paid:bool Description:str
=Items IID:int INUM:int Delivery_Date:date Unit_Price:real Quantity:int Description:str
[#<There is a 1:M relationship between the Invoices and Items tables>
  (Customers
    50 <Best People> <123 Somewhere> <John Doe> <[email protected]> 
    19 <Supersuppliers> ? <Jane Doe> <[email protected]> 
  )
  (Invoices
    152 50 2022-01-17 2022-02-17 no <COD> 
    153 19 2022-01-19 2022-02-19 yes <> 
  )
  (Items
    1839 152 2022-01-16 29.99 2 <Bales of hay> 
    1840 152 2022-01-16 5.98 3 <Straps> 
    1620 153 2022-01-19 11.5 1 <Washers (1-in)> 
  )
]

Here, we've added types to each table's ttype.

It is conventional in a database to have IDs and foreign keys. But these can often be avoided by using hierarchical data. For example:

uxf 1.0 MyApp Data
#<There is a 1:M relationship between the Invoices and Items tables>
=Database customers:Customers invoices:Invoices
=Customers CID:int Company:str Address:str Contact:str Email:str
=Invoices INUM:int CID:int Raised_Date:date Due_Date:date Paid:bool
Description:str Items:Items
=Items IID:int Delivery_Date:date Unit_Price:real Quantity:int Description:str
(Database
    (Customers
    50 <Best People> <123 Somewhere> <John Doe> <[email protected]> 
    19 <Supersuppliers> ? <Jane Doe> <[email protected]> 
    )
    (Invoices
    152 50 2022-01-17 2022-02-17 no <COD> (Items
        1839 2022-01-16 29.99 2 <Bales of hay> 
        1840 2022-01-16 5.98 3 <Straps> 
        )
    153 19 2022-01-19 2022-02-19 yes <> (Items
        1620 2022-01-19 11.5 1 <Washers (1-in)> 
        )
    )
)

Notice that Items no longer need an INUM to identify the Invoice they belong to because they are nested inside their Invoice. However, the relational approach has been retained for Customers since more than one Invoice could be for the same Customer.

In addition, rather than using a simple list of tables, we've created a “Database” ttype and specified it as containing two tables.

What if we wanted to add some extra configuration data to the database? One solution would be to add a third field to the “Database” ttype (e.g., =Database customers:Customers invoices:Invoices config:map). Or we could go further and specify a “Config” ttype and specify the third field as config:Config.

See the testdata folder for more examples of .uxf files (some with other suffixes). See also the t and eg folders in each language-specific library (e.g., py/t and py/eg) for additional examples.

Implementations in additional languages are planned.

UXF files are normally completely self-contained. However, in some cases it may be desirable to share a set of ttype definitions amongst many UXF files.

The disadvantages of doing this are: first, that the relevant UXF files become dependent on one or more external dependencies; second, it is possible to have import conflicts (i.e., two ttypes with the same name but different definitions; and third, if URL imports are used, load times will be affected by network availability and latency. (However, the first and third disadvantages don't apply if all the dependencies are provided by the UXF processor itself, i.e., are system imports.)

The advantage of importing ttype definitions is that for UXF's that have lots of ttypes, only the import(s) and the data need be in the file, without having to repeat all the ttype definitions.

Imports go at the start of the file after the header and after any file-level comment, and before any ttype definitions. Each import must be on its own line and may not span lines, nor have comments.

If a filename import has no path or a relative path, the import attempt will be made relative to the importing .uxf file, and failing that, relative to the current folder, and failing those, relative to each path in the UXF_PATH environment variable (if it exists and is nonempty).

Any ttype definition that follows an import will redefine any imported defintion of the same name.

Imports with no suffix (e.g., complex, fraction, numeric), are provided by the UXF processor itself.

The imported file must be a valid UXF file. It need not have a .uxf suffix (e.g., you might prefer .uxt or .uxi), but must have a suffix (to distinguish it from a system import), and must have a .gz suffix if gzip compressed. Any custom string, comments, or data the imported file may contain are ignored: only the ttype definitions are used.

uxf 1.0
!complex
!fraction
[(Complex 5.1 7.2 8e-2 -9.1e6 0.1 -11.2) <a string> (Fraction 22 7 355 113)]

Here we've used the official system complex's Complex and fraction's Fraction ttypes without having to specify them explicitly. The data represented is a list consisting of three Complex numbers each holding two reals each, a str, and two Fractions holding two ints each.

uxf 1.0
!numeric
[(Complex 5.1 7.2 8e-2 -9.1e6 0.1 -11.2) <a string> (Fraction 22 7 355 113)]

This is the same as the previous example, but using the system convenience numeric import to pull in both the Complex and Fraction ttypes.

If you choose to use imports we recommed that UXF files intended for import either contain a single ttype definition or two or more imports.

We recommend avoiding imports and using stand-alone UXF files wherever possible. Some UXF processors can do UXF to UXF conversions that will replace imports with (actually used) ttype definitions. (For example, the Python UXF library's uxf.py module can do this.)

A UXF file consists of a mandatory header followed by an optional file-level comment, optional imports, optional ttype definitions, and then a single mandatory list, map, or table (which may be empty).

UXF          ::= 'uxf' RWS VERSION CUSTOM? '\n' CONTENT
VERSION      ::= /\d+\.\d+/
CUSTOM       ::= RWS [^\n]+ # user-defined data e.g. filetype and version
CONTENT      ::= COMMENT? IMPORT* TTYPEDEF* (MAP | LIST | TABLE)
IMPORT       ::= '!' /\s*/ IMPORT_FILE '\n' # See below for IMPORT_FILE
TTYPEDEF     ::= '=' COMMENT? OWS IDENFIFIER (RWS FIELD)* # IDENFIFIER is the ttype (i.e., the table name)
FIELD        ::= IDENFIFIER (OWS ':' OWS VALUETYPE)? # IDENFIFIER is the field name
MAP          ::= '{' COMMENT? MAPTYPES? OWS (KEY RWS VALUE)? (RWS KEY RWS VALUE)* OWS '}'
MAPTYPES     ::= OWS KEYTYPE (RWS VALUETYPE)?
KEYTYPE      ::=  'bytes' | 'date' | 'datetime' | 'int' | 'str'
VALUETYPE    ::= KEYTYPE | 'bool' | 'real' | 'list' | 'map' | 'table' | IDENFIFIER # IDENFIFIER is table name
LIST         ::= '[' COMMENT? LISTTYPE? OWS VALUE? (RWS VALUE)* OWS ']'
LISTTYPE     ::= OWS VALUETYPE
TABLE        ::= '(' COMMENT? OWS IDENFIFIER (RWS VALUE)* ')' # IDENFIFIER is the ttype (i.e., the table name)
COMMENT      ::= OWS '#' STR
KEY          ::= BYTES | DATE | DATETIME | INT | STR
VALUE        ::= KEY | NULL | BOOL | REAL | LIST | MAP | TABLE
NULL         ::= '?'
BOOL         ::= 'no' | 'yes'
INT          ::= /[-+]?\d+/
REAL         ::= # standard or scientific notation
DATE         ::= /\d\d\d\d-\d\d-\d\d/ # basic ISO8601 YYYY-MM-DD format
DATETIME     ::= /\d\d\d\d-\d\d-\d\dT\d\d(:\d\d(:\d\d)?)?/ # see note below
STR          ::= /[<][^<>]*?[>]/ # newlines allowed, and &amp; &lt; &gt; supported i.e., XML
BYTES        ::= '(:' (OWS [A-Fa-f0-9]{2})* OWS ':)'
IDENFIFIER   ::= /[_\p{L}]\w{0,59}/ # Must start with a letter or underscore; may not be a built-in typename or constant
OWS          ::= /[\s\n]*/
RWS          ::= /[\s\n]+/ # in some cases RWS is actually optional

Note that a UXF file must contain a single list, map, or table, even if it is empty.

An IMPORT_FILE may be a filename which does not have a file suffix, in which case it is assumed to be a “system” UXF provided by the UXF processor itself. (Currently there are just three system UXFs: complex, fraction, and numeric.) Or it may be a filename with an absolute or relative path. In the latter case the import is searched for in the importing .uxf file's folder, or the current folder, or a folder in the UXF_PATH until it is found—or not). Or it may be a URL referring to an external UXF file. (See Imports.)

To indicate any type valid for the context, simply omit the type name.

As the BNF shows, list, map, and table values may be of any type including nested lists, maps, and tables.

For a table, after the optional comment, there must be an identifier which is the table's ttype. This is followed by the table's values. There's no need to distinguish between one row and the next (although it is common to start new rows on new lines) since the number of fields indicate how many values each row has. It is possible to create tables that have no fields; these might be used for representing enumerations or states.

If a list value, map key, or table value's type is specified, then the UXF processor is expected to be able to check for (and if requested and possible, correct) any mistyped values. UXF readers and writers are expected to preserve map items in the original reading order (first to last, i.e., in insertion order).

For datetime's, only 1-second resolution is supported and no timezones. If microsecond resolution or timezones are required, consider using custom ttypes, e.g.,

=Timestamp when:datetime microseconds:real
=DateTime when:datetime zone:str

Note that a UXF reader (writer) must be able to read (write) a plain text .uxf file containing UTF-8 encoded text, and ought to be able to read and write gzipped plain text .uxf.gz files.

Note also that UXF readers and writers should not care about the actual file extension (apart from the .gz needed for gzipped files), since users are free to use their own. For example, data.myapp and data.myapp.gz.

If you use the vim editor, simple color syntax highlighting is available. Copy uxf.vim into your $VIM/syntax/ folder and add these lines (or similar) to your .vimrc or .gvimrc file:

au BufRead,BufNewFile,BufEnter * if getline(1) =~ '^uxf ' | setlocal ft=uxf | endif
au BufRead,BufNewFile,BufEnter *.uxf set ft=uxf|set expandtab|set tabstop=2|set softtabstop=2|set shiftwidth=2