sea_orm_cli/commands/

generate.rs

use crate::{BannerVersion, BigIntegerType, DateTimeCrate, GenerateSubcommands};
use core::time;
use sea_orm_codegen::{
    BannerVersion as CodegenBannerVersion, BigIntegerType as CodegenBigIntegerType,
    DateTimeCrate as CodegenDateTimeCrate, EntityFormat, EntityTransformer, EntityWriterContext,
    MergeReport, OutputFile, WithPrelude, WithSerde, merge_entity_files,
};
use std::{error::Error, fs, path::Path, process::Command, str::FromStr};
use tracing_subscriber::{EnvFilter, prelude::*};
use url::Url;

/// Split a string by comma while respecting parentheses nesting.
/// This allows attributes like `test(a, b)` to be treated as a single value
/// instead of being split into `test(a` and ` b)`.
fn split_by_comma_ignoring_parentheses(s: &str) -> Vec<String> {
    let mut result = Vec::new();
    let mut current = String::new();
    let mut paren_depth = 0usize;
    let mut bracket_depth = 0usize;
    let mut brace_depth = 0usize;

    for c in s.chars() {
        match c {
            '(' => {
                paren_depth += 1;
                current.push(c);
            }
            ')' => {
                paren_depth = paren_depth.saturating_sub(1);
                current.push(c);
            }
            '[' => {
                bracket_depth += 1;
                current.push(c);
            }
            ']' => {
                bracket_depth = bracket_depth.saturating_sub(1);
                current.push(c);
            }
            '{' => {
                brace_depth += 1;
                current.push(c);
            }
            '}' => {
                brace_depth = brace_depth.saturating_sub(1);
                current.push(c);
            }
            ',' if paren_depth == 0 && bracket_depth == 0 && brace_depth == 0 => {
                let trimmed = current.trim();
                if !trimmed.is_empty() {
                    result.push(trimmed.to_string());
                }
                current.clear();
            }
            _ => {
                current.push(c);
            }
        }
    }

    // Add the last segment
    let trimmed = current.trim();
    if !trimmed.is_empty() {
        result.push(trimmed.to_string());
    }

    result
}

/// Process a vector of strings that may contain comma-separated values with nested parentheses.
/// This handles the case where clap no longer splits by comma, so we need to manually split
/// each string while respecting parentheses nesting.
fn process_comma_separated_values(values: Vec<String>) -> Vec<String> {
    values
        .into_iter()
        .flat_map(|s| split_by_comma_ignoring_parentheses(&s))
        .collect()
}

/// Whether a discovered SQLite column is a generated (computed) column.
///
/// Generated columns cannot be inserted or updated, so they are dropped from
/// generated entities — emitting them as ordinary fields makes every
/// `INSERT`/`UPDATE` fail with "cannot INSERT/UPDATE a generated column" (#3094).
#[cfg(feature = "sqlx-sqlite")]
fn sqlite_column_is_generated(col: &sea_schema::sqlite::def::ColumnInfo) -> bool {
    use sea_schema::sqlite::def::ColumnVisibility;
    matches!(
        col.hidden,
        ColumnVisibility::GeneratedVirtual | ColumnVisibility::GeneratedStored
    )
}

pub async fn run_generate_command(
    command: GenerateSubcommands,
    verbose: bool,
) -> Result<(), Box<dyn Error>> {
    match command {
        GenerateSubcommands::Entity {
            entity_format,
            compact_format: _,
            expanded_format,
            frontend_format,
            include_hidden_tables,
            tables,
            ignore_tables,
            max_connections,
            acquire_timeout,
            output_dir,
            database_schema,
            database_url,
            with_prelude,
            with_serde,
            serde_skip_deserializing_primary_key,
            serde_skip_hidden_column,
            with_copy_enums,
            date_time_crate,
            big_integer_type,
            lib,
            model_extra_derives,
            model_extra_attributes,
            enum_extra_derives,
            enum_extra_attributes,
            column_extra_derives,
            seaography,
            impl_active_model_behavior,
            preserve_user_modifications,
            banner_version,
            er_diagram,
        } => {
            if verbose {
                let _ = tracing_subscriber::fmt()
                    .with_max_level(tracing::Level::DEBUG)
                    .with_test_writer()
                    .try_init();
            } else {
                let filter_layer = EnvFilter::try_new("sea_orm_codegen=info").unwrap();
                let fmt_layer = tracing_subscriber::fmt::layer()
                    .with_target(false)
                    .with_level(false)
                    .without_time();

                let _ = tracing_subscriber::registry()
                    .with(filter_layer)
                    .with(fmt_layer)
                    .try_init();
            }

            // The database should be a valid URL that can be parsed
            // protocol://username:password@host/database_name
            let url = Url::parse(&database_url)?;

            // Make sure we have all the required url components
            //
            // Missing scheme will have been caught by the Url::parse() call
            // above
            let is_sqlite = url.scheme() == "sqlite";

            // Closures for filtering tables
            let filter_tables =
                |table: &String| -> bool { tables.is_empty() || tables.contains(table) };

            let filter_hidden_tables = |table: &str| -> bool {
                if include_hidden_tables {
                    true
                } else {
                    !table.starts_with('_')
                }
            };

            let filter_skip_tables = |table: &String| -> bool { !ignore_tables.contains(table) };

            let _database_name = if !is_sqlite {
                // The database name should be the first element of the path string
                //
                // Throwing an error if there is no database name since it might be
                // accepted by the database without it, while we're looking to dump
                // information from a particular database
                let database_name = url
                    .path_segments()
                    .unwrap_or_else(|| {
                        panic!(
                            "There is no database name as part of the url path: {}",
                            url.as_str()
                        )
                    })
                    .next()
                    .unwrap();

                // An empty string as the database name is also an error
                if database_name.is_empty() {
                    panic!(
                        "There is no database name as part of the url path: {}",
                        url.as_str()
                    );
                }

                database_name
            } else {
                Default::default()
            };

            let (schema_name, table_stmts) = match url.scheme() {
                "mysql" => {
                    #[cfg(not(feature = "sqlx-mysql"))]
                    {
                        panic!("mysql feature is off")
                    }
                    #[cfg(feature = "sqlx-mysql")]
                    {
                        use sea_schema::mysql::discovery::SchemaDiscovery;
                        use sqlx::MySql;

                        println!("Connecting to MySQL ...");
                        let connection = sqlx_connect::<MySql>(
                            max_connections,
                            acquire_timeout,
                            url.as_str(),
                            None,
                        )
                        .await?;
                        println!("Discovering schema ...");
                        let schema_discovery = SchemaDiscovery::new(connection, _database_name);
                        let schema = schema_discovery.discover().await?;
                        let table_stmts = schema
                            .tables
                            .into_iter()
                            .filter(|schema| filter_tables(&schema.info.name))
                            .filter(|schema| filter_hidden_tables(&schema.info.name))
                            .filter(|schema| filter_skip_tables(&schema.info.name))
                            .map(|mut schema| {
                                // Skip generated columns (see #3094).
                                schema.columns.retain(|col| !col.extra.generated);
                                schema.write()
                            })
                            .collect();
                        (None, table_stmts)
                    }
                }
                "sqlite" => {
                    #[cfg(not(feature = "sqlx-sqlite"))]
                    {
                        panic!("sqlite feature is off")
                    }
                    #[cfg(feature = "sqlx-sqlite")]
                    {
                        use sea_schema::sqlite::discovery::SchemaDiscovery;
                        use sqlx::Sqlite;

                        println!("Connecting to SQLite ...");
                        let connection = sqlx_connect::<Sqlite>(
                            max_connections,
                            acquire_timeout,
                            url.as_str(),
                            None,
                        )
                        .await?;
                        println!("Discovering schema ...");
                        let schema_discovery = SchemaDiscovery::new(connection);
                        let schema = schema_discovery
                            .discover()
                            .await?
                            .merge_indexes_into_table();
                        let table_stmts = schema
                            .tables
                            .into_iter()
                            .filter(|schema| filter_tables(&schema.name))
                            .filter(|schema| filter_hidden_tables(&schema.name))
                            .filter(|schema| filter_skip_tables(&schema.name))
                            .map(|mut schema| {
                                // Skip generated columns: codegen can't round-trip them, and
                                // emitting them as ordinary fields makes INSERT/UPDATE fail
                                // ("cannot INSERT/UPDATE a generated column"). See #3094.
                                schema
                                    .columns
                                    .retain(|col| !sqlite_column_is_generated(col));
                                schema.write()
                            })
                            .collect();
                        (None, table_stmts)
                    }
                }
                "postgres" | "postgresql" => {
                    #[cfg(not(feature = "sqlx-postgres"))]
                    {
                        panic!("postgres feature is off")
                    }
                    #[cfg(feature = "sqlx-postgres")]
                    {
                        use sea_schema::postgres::discovery::SchemaDiscovery;
                        use sqlx::Postgres;

                        println!("Connecting to Postgres ...");
                        let schema = database_schema.as_deref().unwrap_or("public");
                        let connection = sqlx_connect::<Postgres>(
                            max_connections,
                            acquire_timeout,
                            url.as_str(),
                            Some(schema),
                        )
                        .await?;
                        println!("Discovering schema ...");
                        let schema_discovery = SchemaDiscovery::new(connection, schema);
                        let schema = schema_discovery.discover().await?;
                        let table_stmts = schema
                            .tables
                            .into_iter()
                            .filter(|schema| filter_tables(&schema.info.name))
                            .filter(|schema| filter_hidden_tables(&schema.info.name))
                            .filter(|schema| filter_skip_tables(&schema.info.name))
                            .map(|mut schema| {
                                // Skip generated columns (see #3094).
                                schema.columns.retain(|col| col.generated.is_none());
                                schema.write()
                            })
                            .collect();
                        (database_schema, table_stmts)
                    }
                }
                _ => unimplemented!("{} is not supported", url.scheme()),
            };
            println!("... discovered.");

            // Process extra derives and attributes, splitting by comma while respecting parentheses
            // This handles cases like `--model-extra-attributes 'cfg_attr(debug_assertions, derive(Debug))'`
            // which should be treated as a single attribute, not split into `cfg_attr(debug_assertions` and ` derive(Debug))`
            let model_extra_derives = process_comma_separated_values(model_extra_derives);
            let model_extra_attributes = process_comma_separated_values(model_extra_attributes);
            let enum_extra_derives = process_comma_separated_values(enum_extra_derives);
            let enum_extra_attributes = process_comma_separated_values(enum_extra_attributes);
            let column_extra_derives = process_comma_separated_values(column_extra_derives);

            let writer_context = EntityWriterContext::new(
                if expanded_format {
                    EntityFormat::Expanded
                } else if frontend_format {
                    EntityFormat::Frontend
                } else if let Some(entity_format) = entity_format {
                    EntityFormat::from_str(&entity_format).expect("Invalid entity-format option")
                } else {
                    EntityFormat::default()
                },
                WithPrelude::from_str(&with_prelude).expect("Invalid prelude option"),
                WithSerde::from_str(&with_serde).expect("Invalid serde derive option"),
                with_copy_enums,
                date_time_crate.into(),
                big_integer_type.into(),
                schema_name,
                lib,
                serde_skip_deserializing_primary_key,
                serde_skip_hidden_column,
                model_extra_derives,
                model_extra_attributes,
                enum_extra_derives,
                enum_extra_attributes,
                column_extra_derives,
                seaography,
                impl_active_model_behavior,
                banner_version.into(),
            );
            let entity_writer = EntityTransformer::transform(table_stmts)?;

            let dir = Path::new(&output_dir);
            fs::create_dir_all(dir)?;

            if er_diagram {
                let diagram = entity_writer.generate_er_diagram();
                let diagram_path = dir.join("entities.mermaid");
                fs::write(&diagram_path, &diagram)?;
                println!("Writing {}", diagram_path.display());
            }

            let output = entity_writer.generate(&writer_context);

            let mut merge_fallback_files: Vec<String> = Vec::new();

            for OutputFile { name, content } in output.files.iter() {
                let file_path = dir.join(name);
                println!("Writing {}", file_path.display());

                if !matches!(
                    name.as_str(),
                    "mod.rs" | "lib.rs" | "prelude.rs" | "sea_orm_active_enums.rs"
                ) && file_path.exists()
                    && preserve_user_modifications
                {
                    let prev_content = fs::read_to_string(&file_path)?;
                    match merge_entity_files(&prev_content, content) {
                        Ok(merged) => {
                            fs::write(file_path, merged)?;
                        }
                        Err(MergeReport {
                            output,
                            warnings,
                            fallback_applied,
                        }) => {
                            for message in warnings {
                                eprintln!("{message}");
                            }
                            fs::write(file_path, output)?;
                            if fallback_applied {
                                merge_fallback_files.push(name.clone());
                            }
                        }
                    }
                } else {
                    fs::write(file_path, content)?;
                };
            }

            // Format each of the files
            for OutputFile { name, .. } in output.files.iter() {
                let exit_status = Command::new("rustfmt").arg(dir.join(name)).status()?; // Get the status code
                if !exit_status.success() {
                    // Propagate the error if any
                    return Err(format!("Fail to format file `{name}`").into());
                }
            }

            if merge_fallback_files.is_empty() {
                println!("... Done.");
            } else {
                return Err(format!(
                    "Merge fallback applied for {} file(s): \n{}",
                    merge_fallback_files.len(),
                    merge_fallback_files.join("\n")
                )
                .into());
            }
        }
    }

    Ok(())
}

async fn sqlx_connect<DB>(
    max_connections: u32,
    acquire_timeout: u64,
    url: &str,
    schema: Option<&str>,
) -> Result<sqlx::Pool<DB>, Box<dyn Error>>
where
    DB: sqlx::Database,
    for<'a> &'a mut <DB as sqlx::Database>::Connection: sqlx::Executor<'a>,
{
    let mut pool_options = sqlx::pool::PoolOptions::<DB>::new()
        .max_connections(max_connections)
        .acquire_timeout(time::Duration::from_secs(acquire_timeout));
    // Set search_path for Postgres, E.g. Some("public") by default
    // MySQL & SQLite connection initialize with schema `None`
    if let Some(schema) = schema {
        let sql = format!("SET search_path = '{schema}'");
        pool_options = pool_options.after_connect(move |conn, _| {
            let sql = sql.clone();
            Box::pin(async move {
                sqlx::Executor::execute(conn, sqlx::AssertSqlSafe(sql))
                    .await
                    .map(|_| ())
            })
        });
    }
    pool_options.connect(url).await.map_err(Into::into)
}

impl From<DateTimeCrate> for CodegenDateTimeCrate {
    fn from(date_time_crate: DateTimeCrate) -> CodegenDateTimeCrate {
        match date_time_crate {
            DateTimeCrate::Chrono => CodegenDateTimeCrate::Chrono,
            DateTimeCrate::Time => CodegenDateTimeCrate::Time,
        }
    }
}

impl From<BigIntegerType> for CodegenBigIntegerType {
    fn from(date_time_crate: BigIntegerType) -> CodegenBigIntegerType {
        match date_time_crate {
            BigIntegerType::I64 => CodegenBigIntegerType::I64,
            BigIntegerType::I32 => CodegenBigIntegerType::I32,
        }
    }
}

impl From<BannerVersion> for CodegenBannerVersion {
    fn from(banner_version: BannerVersion) -> CodegenBannerVersion {
        match banner_version {
            BannerVersion::Off => CodegenBannerVersion::Off,
            BannerVersion::Major => CodegenBannerVersion::Major,
            BannerVersion::Minor => CodegenBannerVersion::Minor,
            BannerVersion::Patch => CodegenBannerVersion::Patch,
        }
    }
}

#[cfg(test)]
mod tests {
    use clap::Parser;

    use super::*;
    use crate::{Cli, Commands};

    #[test]
    #[should_panic(
        expected = "called `Result::unwrap()` on an `Err` value: RelativeUrlWithoutBase"
    )]
    fn test_generate_entity_no_protocol() {
        let cli = Cli::parse_from([
            "sea-orm-cli",
            "generate",
            "entity",
            "--database-url",
            "://root:root@localhost:3306/database",
        ]);

        match cli.command {
            Commands::Generate { command } => {
                smol::block_on(run_generate_command(command, cli.verbose)).unwrap();
            }
            _ => unreachable!(),
        }
    }

    #[test]
    #[should_panic(
        expected = "There is no database name as part of the url path: postgresql://root:root@localhost:3306"
    )]
    fn test_generate_entity_no_database_section() {
        let cli = Cli::parse_from([
            "sea-orm-cli",
            "generate",
            "entity",
            "--database-url",
            "postgresql://root:root@localhost:3306",
        ]);

        match cli.command {
            Commands::Generate { command } => {
                smol::block_on(run_generate_command(command, cli.verbose)).unwrap();
            }
            _ => unreachable!(),
        }
    }

    #[test]
    #[should_panic(
        expected = "There is no database name as part of the url path: mysql://root:root@localhost:3306/"
    )]
    fn test_generate_entity_no_database_path() {
        let cli = Cli::parse_from([
            "sea-orm-cli",
            "generate",
            "entity",
            "--database-url",
            "mysql://root:root@localhost:3306/",
        ]);

        match cli.command {
            Commands::Generate { command } => {
                smol::block_on(run_generate_command(command, cli.verbose)).unwrap();
            }
            _ => unreachable!(),
        }
    }

    #[test]
    #[should_panic(expected = "called `Result::unwrap()` on an `Err` value: EmptyHost")]
    fn test_generate_entity_no_host() {
        let cli = Cli::parse_from([
            "sea-orm-cli",
            "generate",
            "entity",
            "--database-url",
            "postgres://root:root@/database",
        ]);

        match cli.command {
            Commands::Generate { command } => {
                smol::block_on(run_generate_command(command, cli.verbose)).unwrap();
            }
            _ => unreachable!(),
        }
    }

    #[test]
    fn test_split_by_comma_simple() {
        // Simple comma-separated values should split normally
        let result = super::split_by_comma_ignoring_parentheses("a,b,c");
        assert_eq!(result, vec!["a", "b", "c"]);
    }

    #[test]
    fn test_split_by_comma_with_parentheses() {
        // Comma inside parentheses should NOT split
        let result = super::split_by_comma_ignoring_parentheses("test(a, b)");
        assert_eq!(result, vec!["test(a, b)"]);

        // Multiple values, one with parentheses containing comma
        let result = super::split_by_comma_ignoring_parentheses("attr1,test(a, b)");
        assert_eq!(result, vec!["attr1", "test(a, b)"]);
    }

    #[test]
    fn test_split_by_comma_with_nested_parentheses() {
        // Nested parentheses with commas
        let result =
            super::split_by_comma_ignoring_parentheses("cfg_attr(debug_assertions, derive(Debug))");
        assert_eq!(result, vec!["cfg_attr(debug_assertions, derive(Debug))"]);

        // Multiple nested parentheses
        let result = super::split_by_comma_ignoring_parentheses(
            "cfg_attr(feature1, attr(a, b)),cfg_attr(feature2, attr(c, d))",
        );
        assert_eq!(
            result,
            vec![
                "cfg_attr(feature1, attr(a, b))",
                "cfg_attr(feature2, attr(c, d))"
            ]
        );
    }

    #[test]
    fn test_split_by_comma_with_brackets() {
        // Brackets should also be respected
        let result = super::split_by_comma_ignoring_parentheses(
            "serde(rename_all = \"camelCase\"),ts(export)",
        );
        assert_eq!(
            result,
            vec!["serde(rename_all = \"camelCase\")", "ts(export)"]
        );

        // Brackets with commas
        let result = super::split_by_comma_ignoring_parentheses("attr[key, value],other");
        assert_eq!(result, vec!["attr[key, value]", "other"]);
    }

    #[test]
    fn test_split_by_comma_with_braces() {
        // Braces should also be respected
        let result = super::split_by_comma_ignoring_parentheses("derive{a, b},other");
        assert_eq!(result, vec!["derive{a, b}", "other"]);
    }

    #[test]
    fn test_split_by_comma_empty() {
        // Empty string should return empty vec
        let result = super::split_by_comma_ignoring_parentheses("");
        assert!(result.is_empty());

        // Only whitespace should return empty vec
        let result = super::split_by_comma_ignoring_parentheses("   ");
        assert!(result.is_empty());
    }

    #[test]
    fn test_split_by_comma_whitespace_handling() {
        // Whitespace around values should be trimmed
        let result = super::split_by_comma_ignoring_parentheses("  a  ,  b  ");
        assert_eq!(result, vec!["a", "b"]);

        // Whitespace inside parentheses should be preserved
        let result = super::split_by_comma_ignoring_parentheses("test( a , b )");
        assert_eq!(result, vec!["test( a , b )"]);
    }

    #[test]
    fn test_process_comma_separated_values() {
        // Process multiple strings, each potentially containing comma-separated values
        let input = vec![
            "attr1,attr2".to_string(),
            "test(a, b)".to_string(),
            "attr3".to_string(),
        ];
        let result = super::process_comma_separated_values(input);
        assert_eq!(result, vec!["attr1", "attr2", "test(a, b)", "attr3"]);
    }

    #[test]
    fn test_split_by_comma_real_world_examples() {
        // Real-world example: cfg_attr with derive
        let result = super::split_by_comma_ignoring_parentheses(
            "cfg_attr(debug_assertions, derive(Debug)),serde(rename_all = \"camelCase\")",
        );
        assert_eq!(
            result,
            vec![
                "cfg_attr(debug_assertions, derive(Debug))",
                "serde(rename_all = \"camelCase\")"
            ]
        );

        // Real-world example: multiple derives
        let result = super::split_by_comma_ignoring_parentheses(
            "derive(Debug, Clone),derive(Serialize, Deserialize)",
        );
        assert_eq!(
            result,
            vec!["derive(Debug, Clone)", "derive(Serialize, Deserialize)"]
        );
    }

    // Regression test for #3094: generated columns must be dropped during
    // `generate entity`, otherwise they are emitted as ordinary writable fields
    // and every INSERT/UPDATE fails ("cannot INSERT/UPDATE a generated column").
    #[cfg(feature = "sqlx-sqlite")]
    #[test]
    fn test_generate_entity_skips_sqlite_generated_columns() {
        use sea_schema::sea_query::ColumnType;
        use sea_schema::sqlite::def::{ColumnInfo, ColumnVisibility, DefaultType, TableDef};

        let col = |cid, name: &str, hidden| ColumnInfo {
            cid,
            name: name.to_owned(),
            r#type: ColumnType::Integer,
            not_null: true,
            default_value: DefaultType::Unspecified,
            primary_key: cid == 0,
            hidden,
        };

        let mut table = TableDef {
            name: "widget".to_owned(),
            foreign_keys: vec![],
            indexes: vec![],
            constraints: vec![],
            columns: vec![
                col(0, "id", ColumnVisibility::Visible),
                col(1, "w", ColumnVisibility::Visible),
                col(2, "area", ColumnVisibility::GeneratedVirtual),
                col(3, "area_stored", ColumnVisibility::GeneratedStored),
            ],
            auto_increment: false,
        };

        // The predicate flags only the generated columns.
        assert!(!super::sqlite_column_is_generated(&table.columns[0]));
        assert!(!super::sqlite_column_is_generated(&table.columns[1]));
        assert!(super::sqlite_column_is_generated(&table.columns[2]));
        assert!(super::sqlite_column_is_generated(&table.columns[3]));

        // After filtering + write(), generated columns are absent from the DDL.
        table
            .columns
            .retain(|col| !super::sqlite_column_is_generated(col));
        let stmt = table.write();
        let names: Vec<String> = stmt
            .get_columns()
            .iter()
            .map(|c| c.get_column_name())
            .collect();
        assert_eq!(names, ["id", "w"]);
    }
}