Lua Hooks Experimental

As of version 0.86.0, lakeFS supports running hooks without relying on external components using an embedded Lua VM

Using Lua hooks, it is possible to pass a Lua script to be executed directly by the lakeFS server when an action occurs.

The Lua runtime embedded in lakeFS is limited for security reasons. It provides a narrow set of APIs and functions that by default do not allow:

1. Accessing any of the running lakeFS server’s environment
2. Accessing the local filesystem available the lakeFS process

Lua Library reference

aws/s3.get_object(bucket, key)

Returns the body (as a Lua string) of the requested object and a boolean value that is true if the requested object exists

aws/s3.put_object(bucket, key, value)

Sets the object at the given bucket and key to the value of the supplied value string

aws/s3.delete_object(bucket [, key])

Deletes the object at the given key

aws/s3.list_objects(bucket [, prefix, continuation_token, delimiter])

Returns a table of results containing the following structure:

• is_truncated: (boolean) whether there are more results to paginate through using the continuation token
• next_continuation_token: (string) to pass in the next request to get the next page of results
• results (table of tables) information about the objects (and prefixes if a delimiter is used)

a result could in one of the following structures

{
   ["key"] = "a/common/prefix/",
   ["type"] = "prefix"
}

or:

{
   ["key"] = "path/to/object",
   ["type"] = "object",
   ["etag"] = "etagString",
   ["size"] = 1024,
   ["last_modified"] = "2023-12-31T23:10:00Z"
}

aws/s3.delete_recursive(bucket, prefix)

Deletes all objects under the given prefix

crypto/aes/encryptCBC(key, plaintext)

Returns a ciphertext for the aes encrypted text

crypto/aes/decryptCBC(key, ciphertext)

Returns the decrypted (plaintext) string for the encrypted ciphertext

crypto/hmac/sign_sha256(message, key)

Returns a SHA256 hmac signature for the given message with the supplied key (using the SHA256 hashing algorithm)

crypto/hmac/sign_sha1(message, key)

Returns a SHA1 hmac signature for the given message with the supplied key (using the SHA1 hashing algorithm)

crypto/md5/digest(data)

Returns the MD5 digest (string) of the given data

crypto/sha256/digest(data)

Returns the SHA256 digest (string) of the given data

encoding/base64/encode(data)

Encodes the given data to a base64 string

encoding/base64/decode(data)

Decodes the given base64 encoded data and return it as a string

encoding/base64/url_encode(data)

Encodes the given data to an unpadded alternate base64 encoding defined in RFC 4648.

encoding/base64/url_decode(data)

Decodes the given unpadded alternate base64 encoding defined in RFC 4648 and return it as a string

encoding/hex/encode(value)

Encode the given value string to hexadecimal values (string)

encoding/hex/decode(value)

Decode the given hexadecimal string back to the string it represents (UTF-8)

encoding/json/marshal(table)

Encodes the given table into a JSON string

encoding/json/unmarshal(string)

Decodes the given string into the equivalent Lua structure

encoding/parquet/get_schema(payload)

Read the payload (string) as the contents of a Parquet file and return its schema in the following table structure:

{
  { ["name"] = "column_a", ["type"] = "INT32" },
  { ["name"] = "column_b", ["type"] = "BYTE_ARRAY" }
}

lakefs

The Lua Hook library allows calling back to the lakeFS API using the identity of the user that triggered the action. For example, if user A tries to commit and triggers a pre-commit hook - any call made inside that hook to the lakeFS API, will automatically use user A’s identity for authorization and auditing purposes.

lakefs/create_tag(repository_id, reference_id, tag_id)

Create a new tag for the given reference

lakefs/diff_refs(repository_id, lef_reference_id, right_reference_id [, after, prefix, delimiter, amount])

Returns an object-wise diff between left_reference_id and right_reference_id.

lakefs/list_objects(repository_id, reference_id [, after, prefix, delimiter, amount])

List objects in the specified repository and reference (branch, tag, commit ID, etc.). If delimiter is empty, will default to a recursive listing. Otherwise, common prefixes up to delimiter will be shown as a single entry.

lakefs/get_object(repository_id, reference_id, path)

Returns 2 values:

1. The HTTP status code returned by the lakeFS API
2. The content of the specified object as a lua string

path/parse(path_string)

Returns a table for the given path string with the following structure:

> require("path")
> path.parse("a/b/c.csv")
{
    ["parent"] = "a/b/"
    ["base_name"] = "c.csv"
} 

path/join(*path_parts)

Receives a variable number of strings and returns a joined string that represents a path:

> require("path")
> path.join("path/", "to", "a", "file.data")
path/o/a/file.data

path/is_hidden(path_string [, seperator, prefix])

returns a boolean - true if the given path string is hidden (meaning it starts with prefix) - or if any of its parents start with prefix.

> require("path")
> path.is_hidden("a/b/c") -- false
> path.is_hidden("a/b/_c") -- true
> path.is_hidden("a/_b/c") -- true
> path.is_hidden("a/b/_c/") -- true

path/default_separator()

Returns a constant string (/)

regexp/match(pattern, s)

Returns true if the string s matches pattern. This is a thin wrapper over Go’s regexp.MatchString.

regexp/quote_meta(s)

Escapes any meta-characters in string s and returns a new string

regexp/compile(pattern)

Returns a regexp match object for the given pattern

regexp/compiled_pattern.find_all(s, n)

Returns a table list of all matches for the pattern, (up to n matches, unless n == -1 in which case all possible matches will be returned)

regexp/compiled_pattern.find_all_submatch(s, n)

Returns a table list of all sub-matches for the pattern, (up to n matches, unless n == -1 in which case all possible matches will be returned). Submatches are matches of parenthesized subexpressions (also known as capturing groups) within the regular expression, numbered from left to right in order of opening parenthesis. Submatch 0 is the match of the entire expression, submatch 1 is the match of the first parenthesized subexpression, and so on

regexp/compiled_pattern.find(s)

Returns a string representing the left-most match for the given pattern in string s

regexp/compiled_pattern.find_submatch(s)

find_submatch returns a table of strings holding the text of the leftmost match of the regular expression in s and the matches, if any, of its submatches

strings/split(s, sep)

returns a table of strings, the result of splitting s with sep.

strings/trim(s)

Returns a string with all leading and trailing white space removed, as defined by Unicode

strings/replace(s, old, new, n)

Returns a copy of the string s with the first n non-overlapping instances of old replaced by new. If old is empty, it matches at the beginning of the string and after each UTF-8 sequence, yielding up to k+1 replacements for a k-rune string.

If n < 0, there is no limit on the number of replacements

strings/has_prefix(s, prefix)

Returns true if s begins with prefix

strings/has_suffix(s, suffix)

Returns true if s ends with suffix

strings/contains(s, substr)

Returns true if substr is contained anywhere in s

time/now()

Returns a float64 representing the amount of nanoseconds since the unix epoch (01/01/1970 00:00:00).

time/format(epoch_nano, layout, zone)

Returns a string representation of the given epoch_nano timestamp for the given Timezone (e.g. "UTC", "America/Los_Angeles", …) The layout parameter should follow Go’s time layout format.

time/format_iso(epoch_nano, zone)

Returns a string representation of the given epoch_nano timestamp for the given Timezone (e.g. "UTC", "America/Los_Angeles", …) The returned string will be in ISO8601 format.

time/sleep(duration_ns)

Sleep for duration_ns nanoseconds

time/since(epoch_nano)

Returns the amount of nanoseconds elapsed since epoch_nano

time/add(epoch_time, duration_table)

Returns a new timestamp (in nanoseconds passed since 01/01/1970 00:00:00) for the given duration. The duration should be a table with the following structure:

> require("time")
> time.add(time.now(), {
    ["hour"] = 1,
    ["minute"] = 20,
    ["second"] = 50
})

You may omit any of the fields from the table, resulting in a default value of 0 for omitted fields

time/parse(layout, value)

Returns a float64 representing the amount of nanoseconds since the unix epoch (01/01/1970 00:00:00). This timestamp will represent date value parsed using the layout format.

The layout parameter should follow Go’s time layout format

time/parse_iso(value)

Returns a float64 representing the amount of nanoseconds since the unix epoch (01/01/1970 00:00:00 for value. The value string should be in ISO8601 format

uuid/new()

Returns a new 128-bit RFC 4122 UUID in string representation.

net/http (optional)

Provides a request function that performs an HTTP request. For security reasons, this package is not available by default as it enables http requests to be sent out from the lakeFS instance network. The feature should be enabled under actions.lua.net_http_enabled configuration. Request will time out after 30 seconds.

http.request(url [, body])
http.request{
  url = string,
  [method = string,]
  [headers = header-table,]
  [body = string,]
}

Returns a code (number), body (string), headers (table) and status (string).

• code - status code number
• body - string with the response body
• headers - table with the response request headers (key/value or table of values)
• status - status code text

The first form of the call will perform GET requests or POST requests if the body parameter is passed.

The second form accepts a table and allows you to customize the request method and headers.

Example of a GET request

local http = require("net/http")
local code, body = http.request("https://example.com")
if code == 200 then
    print(body)
else
    print("Failed to get example.com - status code: " .. code)
end

Example of a POST request

local http = require("net/http")
local code, body = http.request{
    url="https://httpbin.org/post",
    method="POST",
    body="custname=tester",
    headers={["Content-Type"]="application/x-www-form-urlencoded"},
}
if code == 200 then
    print(body)
else
    print("Failed to post data - status code: " .. code)
end

Example Hooks

This example will simply print out a JSON representation of the event that occurred:

name: dump_all
on:
  post-commit:
  post-merge:
  post-create-tag:
  post-create-branch:
hooks:
  - id: dump_event
    type: lua
    properties:
      script: |
        json = require("encoding/json")
        print(json.marshal(action))

A more useful example: ensure every commit contains a required metadata field:

name: pre commit metadata field check
on:
pre-commit:
    branches:
    - main
    - dev
hooks:
  - id: ensure_commit_metadata
    type: lua
    properties:
      args:
        notebook_url: {"pattern": "my-jupyter.example.com/.*"}
        spark_version:  {}
      script: |
        regexp = require("regexp")
        for k, props in pairs(args) do
          current_value = action.commit.metadata[k]
          if current_value == nil then
            error("missing mandatory metadata field: " .. k)
          end
          if props.pattern and not regexp.match(props.pattern, current_value) then
            error("current value for commit metadata field " .. k .. " does not match pattern: " .. props.pattern .. " - got: " .. current_value)
          end
        end

For more examples and configuration samples, check out the examples/hooks/ directory in the lakeFS repository.