This is part 22 of a series on how to approach bash programming in a way that’s safer and more structured than your basic script.

See part 1 if you want to catch the series from the start.

Last time, we discussed environment variables. This time, let’s discuss data types.

Basic Types

Bash is like many dynamic languages in that it has a variety of data types which include collection types, and does not require declaration of types before assigning values to variables (with the exception of hashes).

For basic types, bash supports two types:

  • strings - ‘natch

  • integers - you know, for math

Strings can actually be used as integers as well, that is to say, they work fine in arithmetic expressions so long as they hold the string representations of integers.

Although it’s fairly easy to use a string as an integer, declaring a variable as an integer does get you something. It allows you use an arithmetic expression as a string on the right-hand side of an assignment, without the need for an arithmetic expansion.

For example, you can use the assignment local -i var=’2 + 5’ to give var the value of 7. Once a variable has the integer attribute, you can do the same without the local or declare attached.

The integer attribute also allows you to use the += operator as a mathematical addition, rather than appending a string.

It also happens to convert any non-integer string values you attempt to assign to the variable into the value 0 (which can hardly be described as a feature). So be careful.

I use the integer declaration when it saves me some of the noisy arithmetic expansion syntax. Fewer $(())s are better.

Strings are the default, so they don’t need any special option to declare or local.

Collection Types

There are two collection types:

  • arrays - start-at-zero-indexed dynamic lists

  • associative arrays - a.k.a. hashes - key-value pairs of the basic types

Despite the fact that bash refers to both of these types as arrays, I prefer to keep things clear by referring to one as arrays and the other as hashes exclusively.

Arrays can skip indexes or have items deleted, which makes them sparse.

It is not possible to store collection types as items in another collection.

Assignment to individual array items uses bracket notation, e.g. myarray[0]=zero. The index can be an arithmetic expression following the usual rules for such expressions (dollar-signs not needed for variable references, etc.).

Hash keys are strings which can include spaces and do not require quotes. Like arrays, the values can be a mix of strings and integers. Hashes do not interpret keys as arithmetic expressions.

Both arrays and hashes can be declared with the integer option, which forces all elements to automatically be an integer. You cannot mix strings and integer types within an array nor hash.

Other Attributes

Variables can have other attributes assigned in their declare or local statements:

  • read-only - declare -r or readonly - make the variable an unchangeable constant

  • global - declare -g - (local doesn’t support -g) force the variable to the global scope

  • export - declare -x or export - export a variable to the environment

Assigning Values

Assigning values to the basic types is straightforward so I won’t go over it. declare and local also allow you to assign values in the same manner. They also allow assignments to multiple variables in one statement. You can do the same with regular assignments as well.

There are two forms for assigning values to the collection types.

The first is assignment to an individual element in the collection, which follows the same syntax as assigning to a regular variable, but includes an index/key on the left-hand side:

myhash[key]=myvalue
myarray[index]=myvalue

For a hash, you must declare the variable with declare -A before assigning values this way.

For arrays, assigning to the unsubscripted name is the same as assigning to the zeroth element. Expanding the unsubscripted name also expands the zeroth element.

The second form is the assignment of a literal to the name of the hash or array. For arrays, it looks like:

myarray=( value1 value2 "value 3" )

Arrays can also have their indexes supplied if you want them to be sparse:

myarray=( [0]=value1 [2]=value3 )

For hashes, it looks like:

declare -A myhash=( [one]=value1 [two]=value2 [three]="value 3" )

Unlike the other types, hashes must be declared as such before or during assignment with the declare -A or local -A command.

Once declared, you can assign its value without redeclaring it though:

declare -A myhash
myhash=( [zero]=0 [one]=1 )

The printf command can also assign directly to array/hash elements (as well as regular variables) with the -v option.

Arrays also support the += operator, which appends to the end:

myarray+=( "new tail value" )

Unsetting an index of an array or hash removes that one element:

unset -v myarray[3]

For this reason, when looping through indexes of an array (or hash, for that matter), you should use the index expansion rather than a counter (reminder: IFS is set to blank):

for index in ${!myarray[@]}; do
  echo ${myarray[index]}
done

Namerefs

Finally, namerefs are a special type of variable in recent versions of bash. There are a form of indirection:

declare -n myvar=variable_name

After the declaration, any reference to myvar will actually read or write to the variable variable_name instead.

Automatic Conversion

You can change a string variable to an array automatically if you use the literal assignment syntax to assign an array value to an existing string variable, or if you assign a value to a subscript of an existing string variable.

Trying the same thing with a hash literal assignment, will turn the variable into an array, not a hash. There’s no way to get around having to use declare -A to get a hash.

Once a variable has been turned into an array type, it cannot be converted to one of the basic types without unsetting it explicitly.

Since the conversion is irreversible, I try not to convert a string to an array so I always know what kind of variable I’m dealing with.

Continue with part 22.5 - naming and namespaces