UPLOAD

Mark endpoint as a file upload handler.

Keywords

@upload, upload

Syntax

@upload
@upload for <handler_type>

If no handler is specified (only upload annotation without for), then the default handler will be used. The default handler is large_object unless configured otherwise via DefaultUploadHandler setting.

Handler Types

There are 4 handler types available:

Handler	Key	Description
Large Object	`large_object`	Stores files using PostgreSQL Large Objects API (default)
File System	`file_system`	Stores files on the server file system
CSV	`csv`	Parses CSV files and processes rows via PostgreSQL command
Excel	`excel`	Parses Excel files and processes rows via PostgreSQL command

Shared Annotation Options

These options are available for all handler types:

Option	Type	Default	Description
`stop_after_first_success`	bool	`false`	Stop upload after first successful upload when multiple handlers are used. Subsequent files will have status `Ignored`.
`included_mime_types`	string	`null`	CSV string of MIME type patterns to include. Set to `null` to allow all.
`excluded_mime_types`	string	`null`	CSV string of MIME type patterns to exclude. Set to `null` to exclude none.
`buffer_size`	int	`null`	Buffer size in bytes for raw content uploads (`large_object` and `file_system`).
`check_text`	bool	`false`	Validate file is a text file (not binary). Set to `true` to accept only text files.
`check_image`	bool/string	`false`	Validate file is an image. Set to `true` to accept only images, or CSV of allowed types: `jpg`, `png`, `gif`, `bmp`, `tiff`, `webp`.
`test_buffer_size`	int	`4096`	Buffer size in bytes when checking text files.
`non_printable_threshold`	int	`5`	Maximum non-printable characters allowed in test buffer to consider a valid text file.

Upload Metadata

All handlers return upload metadata as JSON with the following common properties:

Property	Type	Description
`type`	string	Handler type used (`large_object`, `file_system`, `csv`, `excel`)
`fileName`	string	Original uploaded file name
`contentType`	string	MIME type of the uploaded file
`size`	int	File size in bytes
`success`	bool	Whether the upload succeeded
`status`	string	Status message (e.g., `Ok`, `InvalidMimeType`)

Handler-specific properties:

Handler	Property	Type	Description
`large_object`	`oid`	int	PostgreSQL Large Object OID
`file_system`	`filePath`	string	Path where file was saved

Large Object Handler

Default handler that stores files using PostgreSQL Large Objects.

Basic Example

sql

create function lo_simple_upload(
    _meta json = null
)
returns json
language plpgsql
as
$$
begin
    return _meta;
end;
$$;

comment on function lo_simple_upload(json) is '
@upload
@param _meta is upload metadata
';

With Custom OID Parameter

You can specify a custom OID for the large object:

sql

create function lo_custom_parameter_upload(
    _oid bigint,
    _meta json = null
)
returns json
language plpgsql
as
$$
begin
    return _meta;
end;
$$;

comment on function lo_custom_parameter_upload(bigint, json) is '
@upload for large_object
@param _meta is upload metadata
@oid = {_oid}
';

Context Metadata

Upload metadata is also available via PostgreSQL context setting:

sql

create function lo_simple_upload_context_metadata()
returns json
language plpgsql
as
$$
begin
    return current_setting('request.upload_metadata', true)::text;
end;
$$;

comment on function lo_simple_upload_context_metadata() is '@upload';

Large Object Annotation Options

All shared options plus:

Option	Description
`oid`	Custom OID for the large object
`large_object_included_mime_types`	Handler-specific MIME types to include
`large_object_excluded_mime_types`	Handler-specific MIME types to exclude
`large_object_buffer_size`	Handler-specific buffer size
`large_object_oid`	Handler-specific OID (alias for `oid`)
`large_object_check_text`	Handler-specific text check
`large_object_check_image`	Handler-specific image check
`large_object_test_buffer_size`	Handler-specific test buffer size
`large_object_non_printable_threshold`	Handler-specific non-printable threshold

File System Handler

Stores files on the server file system.

Basic Example

sql

create function fs_simple_upload(
    _meta json = null
)
returns json
language plpgsql
as
$$
begin
    return _meta;
end;
$$;

comment on function fs_simple_upload(json) is '
@upload for file_system
@param _meta is upload metadata
';

With Custom Parameters

Control the file path, name, and behavior:

sql

create function fs_custom_parameter_upload(
    _path text,
    _file text,
    _unique_name boolean,
    _create_path boolean,
    _meta json = null
)
returns json
language plpgsql
as
$$
begin
    return _meta;
end;
$$;

comment on function fs_custom_parameter_upload(text, text, boolean, boolean, json) is '
@upload for file_system
@param _meta is upload metadata
@path = {_path}
@file = {_file}
@unique_name = {_unique_name}
@create_path = {_create_path}
';

File System Annotation Options

All shared options plus:

Option	Description
`path`	Directory path for uploaded file
`file`	File name to use
`unique_name`	Generate unique file name (bool)
`create_path`	Create directory if not exists (bool)
`file_system_included_mime_types`	Handler-specific MIME types to include
`file_system_excluded_mime_types`	Handler-specific MIME types to exclude
`file_system_buffer_size`	Handler-specific buffer size
`file_system_path`	Handler-specific path (alias for `path`)
`file_system_file`	Handler-specific file name (alias for `file`)
`file_system_unique_name`	Handler-specific unique name setting
`file_system_create_path`	Handler-specific create path setting
`file_system_check_text`	Handler-specific text check
`file_system_check_image`	Handler-specific image check
`file_system_test_buffer_size`	Handler-specific test buffer size
`file_system_non_printable_threshold`	Handler-specific non-printable threshold

MIME Type Filtering

sql

comment on function fs_upload_include_mime_type(json) is '
@upload for file_system
@param _meta is upload metadata
@path = ./test
@file = mime_type.csv
@included_mime_types = image/*, application/*
';

CSV Handler

Parses CSV files and processes each row via a PostgreSQL command.

Row Command Function Signature

The row command function receives up to 4 parameters:

sql

create function my_csv_row_processor(
    _index int,           -- $1: Row index (1-based)
    _row text[],          -- $2: Parsed row values as text array
    _prev_result any,     -- $3: Result of previous row command (for chaining)
    _meta json            -- $4: Row metadata JSON
)
returns any               -- Return value passed to next row as $3

Row Command Parameters

Parameter	Type	Description
`$1`	`int`	Row index (1-based, includes header row)
`$2`	`text[]`	Parsed row values as text array (e.g., `_row[1]`, `_row[2]`, etc.)
`$3`	`any`	Result of previous row command execution (see below)
`$4`	`json`	Row metadata JSON object

Row chaining with $3: The return value from each row command is passed to the next row as $3. For the first row, $3 is NULL. If the row command returns void (no return value), $3 will be NULL for the next row. This enables accumulating values across rows (e.g., counting rows, summing values).

Row Metadata Structure (`$4`)

The metadata JSON passed to each row command contains:

json

{
  "type": "csv",
  "fileName": "data.csv",
  "contentType": "text/csv",
  "size": 1234,
  "claims": {                    // Only if RowCommandUserClaimsKey is set (default: "claims")
    "user_id": "1",
    "user_name": "alice",
    "name_identifier": "1"
  }
}

Property	Type	Description
`type`	string	Handler type (`"csv"`)
`fileName`	string	Original uploaded file name
`contentType`	string	MIME type of the file
`size`	int	File size in bytes
`claims`	object	User claims (when `RowCommandUserClaimsKey` is configured)

Note: Unlike Excel, CSV row metadata does NOT include rowIndex. Use the $1 parameter for the row index.

Upload Function Metadata (`_meta` parameter)

The main upload function receives metadata as a JSON array with one element per uploaded file:

json

[
  {
    "type": "csv",
    "fileName": "data.csv",
    "contentType": "text/csv",
    "size": 1234,
    "success": true,
    "status": "Ok",
    "lastResult": 100
  }
]

Property	Type	Description
`lastResult`	any	Final return value from the last row command execution

Basic Example

sql

-- Table for uploads
create table csv_uploads (
    id int primary key generated always as identity,
    file_name text not null,
    row_index int not null,
    row_data text[] not null
);

-- Row command to process each CSV row
create function csv_upload_row(
    _index int,
    _row text[],
    _prev_result int,
    _meta json
)
returns int
language plpgsql
as $$
begin
    insert into csv_uploads (file_name, row_index, row_data)
    values (_meta->>'fileName', _index, _row);

    return coalesce(_prev_result, 0) + 1;
end;
$$;

-- HTTP POST endpoint
create function csv_upload(_meta json = null)
returns json
language sql
as $$
    select _meta;
$$;

comment on function csv_upload(json) is '
@upload for csv
@param _meta is upload metadata
@row_command = select csv_upload_row($1,$2,$3,$4)
';

Accessing User Claims in Row Command

With RowCommandUserClaimsKey configured (default: "claims"), user claims are available in the row metadata:

sql

create function csv_upload_row(
    _index int,
    _row text[],
    _prev_result int,
    _meta json
)
returns int
language plpgsql
as $$
begin
    insert into csv_uploads (user_id, file_name, row_index, row_data)
    values (
        (_meta->'claims'->>'user_id')::int,  -- Access user_id from claims
        _meta->>'fileName',
        _index,
        _row
    );
    return coalesce(_prev_result, 0) + 1;
end;
$$;

Using User Context Variables

With UseUserContext: true, user context variables are set before upload and accessible via current_setting():

sql

create function csv_upload_row(
    _index int,
    _row text[],
    _prev_result int,
    _meta json
)
returns int
language plpgsql
as $$
begin
    insert into csv_uploads (user_id, file_name, row_index, row_data)
    values (
        current_setting('request.user_id')::int,  -- Access from context
        _meta->>'fileName',
        _index,
        _row
    );
    return coalesce(_prev_result, 0) + 1;
end;
$$;

Custom Delimiters

Support multiple delimiter characters:

sql

comment on function csv_upload(json) is '
@upload for csv
@param _meta is upload metadata
@delimiters = ,;
@row_command = select csv_upload_row($1,$2,$3,$4)
';

This will use comma (,) and semicolon (;) as delimiters. Use \t for tab.

CSV Annotation Options

All shared options (except buffer_size, check_text, check_image) plus:

Option	Type	Default	Description
`row_command`	string	-	PostgreSQL command to process each row (required)
`delimiters`	string	`,`	Delimiter character(s)
`check_format`	bool	`false`	Validate file is text before processing
`has_fields_enclosed_in_quotes`	bool	`true`	Fields may be enclosed in quotes
`set_white_space_to_null`	bool	`true`	Convert whitespace-only values to NULL

Handler-specific prefixed aliases are also available (e.g., csv_row_command, csv_delimiters).

Excel Handler

Parses Excel files (.xlsx, .xls) and processes each row via a PostgreSQL command.

Row Command Function Signature

The row command function receives up to 4 parameters:

sql

create function my_excel_row_processor(
    _index int,           -- $1: Row index (1-based, non-empty rows only)
    _row text[],          -- $2: Row values as text array (or json if row_is_json = true)
    _prev_result any,     -- $3: Result of previous row command
    _meta json            -- $4: Row metadata JSON (includes sheet name)
)
returns any               -- Return value passed to next row as $3

Row Command Parameters

Parameter	Type	Description
`$1`	`int`	Row index (1-based, only counts non-empty rows)
`$2`	`text[]` or `json`	Row values as text array, or JSON if `row_is_json = true`
`$3`	`any`	Result of previous row command execution (see below)
`$4`	`json`	Row metadata JSON object (includes sheet info)

Row Metadata Structure (`$4`)

The metadata JSON passed to each row command contains:

json

{
  "type": "excel",
  "fileName": "data.xlsx",
  "contentType": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
  "size": 5678,
  "sheet": "Sheet1",
  "rowIndex": 5,
  "claims": {                    // Only if RowCommandUserClaimsKey is set
    "user_id": "1",
    "user_name": "alice"
  }
}

Property	Type	Description
`type`	string	Handler type (`"excel"`)
`fileName`	string	Original uploaded file name
`contentType`	string	MIME type of the file
`size`	int	File size in bytes
`sheet`	string	Current sheet name being processed
`rowIndex`	int	Excel row index (1-based, includes empty rows)
`claims`	object	User claims (when `RowCommandUserClaimsKey` is configured)

Note: Excel row metadata includes rowIndex (actual Excel row number) and sheet name. The $1 parameter is a sequential counter for non-empty rows only, while rowIndex reflects the actual Excel row position.

Upload Function Metadata (`_meta` parameter)

The main upload function receives metadata as a JSON array. When all_sheets = true, there's one element per sheet:

json

[
  {
    "type": "excel",
    "fileName": "data.xlsx",
    "contentType": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
    "size": 5678,
    "sheet": "Sheet1",
    "success": true,
    "rows": 99,
    "result": 100
  },
  {
    "type": "excel",
    "fileName": "data.xlsx",
    "contentType": "...",
    "size": 5678,
    "sheet": "Sheet2",
    "success": true,
    "rows": 50,
    "result": 50
  }
]

Property	Type	Description
`sheet`	string	Sheet name
`rows`	int	Number of non-empty rows processed
`result`	any	Final return value from the last row command for this sheet

Basic Example

sql

-- Table for uploads
create table excel_uploads (
    id int primary key generated always as identity,
    file_name text not null,
    sheet_name text,
    row_index int not null,
    row_data text[] not null
);

-- Row command to process each Excel row
create function excel_upload_row(
    _index int,
    _row text[],
    _prev_result int,
    _meta json
)
returns int
language plpgsql
as $$
begin
    insert into excel_uploads (file_name, sheet_name, row_index, row_data)
    values (
        _meta->>'fileName',
        _meta->>'sheet',
        _index,
        coalesce(_row, '{}')
    );

    return coalesce(_prev_result, 0) + 1;
end;
$$;

-- HTTP POST endpoint
create function excel_upload(_meta json = null)
returns json
language sql
as $$
    select _meta;
$$;

comment on function excel_upload(json) is '
@upload for excel
@param _meta is upload metadata
@all_sheets = true
@row_command = select excel_upload_row($1,$2,$3,$4)
';

Row Data as JSON

When row_is_json = true, row data is passed as JSON with Excel cell references as keys:

sql

comment on function excel_upload(json) is '
@upload for excel
@param _meta is upload metadata
@row_is_json = true
@row_command = select excel_upload_row($1,$2,$3,$4)
';

The $2 parameter becomes JSON like:

json

{"A1": "Name", "B1": "Value", "C1": 123}

Excel Annotation Options

All shared options (except buffer_size, check_text, check_image, test_buffer_size, non_printable_threshold) plus:

Option	Type	Default	Description
`row_command`	string	-	PostgreSQL command to process each row (required)
`sheet_name`	string	`null`	Specific sheet name to process (first sheet if null)
`all_sheets`	bool	`false`	Process all sheets in the workbook
`time_format`	string	`HH:mm:ss`	Format for time values
`date_format`	string	`yyyy-MM-dd`	Format for date values
`datetime_format`	string	`yyyy-MM-dd HH:mm:ss`	Format for datetime values
`row_is_json`	bool	`false`	Pass row data as JSON instead of text array

Handler-specific prefixed aliases are also available (e.g., excel_row_command, excel_all_sheets).

Error Handling and Rollback

All upload handlers support automatic rollback on error. If the handler function raises an exception, any uploaded data is rolled back:

sql

create function lo_upload_raise_exception(
    _oid bigint,
    _meta json = null
)
returns json
language plpgsql
as
$$
begin
    raise exception 'failed upload';
    return _meta;
end;
$$;

comment on function lo_upload_raise_exception(bigint, json) is '
@upload for large_object
@param _meta is upload metadata
@oid = {_oid}
';

If an exception is raised:

Large Object: The large object is deleted
File System: The uploaded file is deleted
CSV/Excel: All database changes are rolled back

Multiple File Uploads

Upload endpoints support multiple files in a single request. The metadata will be returned as a JSON array with one entry per file.

Behavior

Enables multipart/form-data file uploads
Handlers process the uploaded file (storage, validation, parsing)
Metadata parameter receives file information (name, size, type)
All uploads are transactional - errors trigger rollback
See Upload Options for configuration

Upload configuration - Configure upload handlers and settings
Comment Annotations Guide - How annotations work
Configuration Guide - How configuration works

Blog Posts

CSV and Excel Ingestion Made Easy - Complete tutorial on CSV/Excel row processing
Secure Image Uploads - Image uploads with Large Objects and File System handlers

AUTHORIZE - Protect upload endpoint

UPLOAD ​

Keywords ​

Syntax ​

Handler Types ​

Shared Annotation Options ​

Upload Metadata ​

Large Object Handler ​

Basic Example ​

With Custom OID Parameter ​

Context Metadata ​

Large Object Annotation Options ​

File System Handler ​

Basic Example ​

With Custom Parameters ​

File System Annotation Options ​

MIME Type Filtering ​

CSV Handler ​

Row Command Function Signature ​

Row Command Parameters ​

Row Metadata Structure ($4) ​

Upload Function Metadata (_meta parameter) ​

Basic Example ​

Accessing User Claims in Row Command ​

Using User Context Variables ​

Custom Delimiters ​

CSV Annotation Options ​

Excel Handler ​

Row Command Function Signature ​

Row Command Parameters ​

Row Metadata Structure ($4) ​

Upload Function Metadata (_meta parameter) ​

Basic Example ​

Row Data as JSON ​

Excel Annotation Options ​

Error Handling and Rollback ​

Multiple File Uploads ​

Behavior ​

Related ​

Blog Posts ​

Related Annotations ​

Comments

UPLOAD

Keywords

Syntax

Handler Types

Shared Annotation Options

Upload Metadata

Large Object Handler

Basic Example

With Custom OID Parameter

Context Metadata

Large Object Annotation Options

File System Handler

Basic Example

With Custom Parameters

File System Annotation Options

MIME Type Filtering

CSV Handler

Row Command Function Signature

Row Command Parameters

Row Metadata Structure (`$4`)

Upload Function Metadata (`_meta` parameter)

Basic Example

Accessing User Claims in Row Command

Using User Context Variables

Custom Delimiters

CSV Annotation Options

Excel Handler

Row Command Function Signature

Row Command Parameters

Row Metadata Structure (`$4`)

Upload Function Metadata (`_meta` parameter)

Basic Example

Row Data as JSON

Excel Annotation Options

Error Handling and Rollback

Multiple File Uploads

Behavior

Related

Blog Posts

Related Annotations