ext/wasm/api/sqlite3-api-opfs.js

/*
  2022-09-18

  The author disclaims copyright to this source code.  In place of a
  legal notice, here is a blessing:

  *   May you do good and not evil.
  *   May you find forgiveness for yourself and forgive others.
  *   May you share freely, never taking more than you give.

  ***********************************************************************

  This file holds the synchronous half of an sqlite3_vfs
  implementation which proxies, in a synchronous fashion, the
  asynchronous Origin-Private FileSystem (OPFS) APIs using a second
  Worker, implemented in sqlite3-opfs-async-proxy.js.  This file is
  intended to be appended to the main sqlite3 JS deliverable somewhere
  after sqlite3-api-oo1.js and before sqlite3-api-cleanup.js.
*/
'use strict';
self.sqlite3ApiBootstrap.initializers.push(function(sqlite3){
/**
   installOpfsVfs() returns a Promise which, on success, installs an
   sqlite3_vfs named "opfs", suitable for use with all sqlite3 APIs
   which accept a VFS. It is intended to be called via
   sqlite3ApiBootstrap.initializersAsync or an equivalent mechanism.

   The installed VFS uses the Origin-Private FileSystem API for
   all file storage. On error it is rejected with an exception
   explaining the problem. Reasons for rejection include, but are
   not limited to:

   - The counterpart Worker (see below) could not be loaded.

   - The environment does not support OPFS. That includes when
     this function is called from the main window thread.

  Significant notes and limitations:

  - As of this writing, OPFS is still very much in flux and only
    available in bleeding-edge versions of Chrome (v102+, noting that
    that number will increase as the OPFS API matures).

  - The OPFS features used here are only available in dedicated Worker
    threads. This file tries to detect that case, resulting in a
    rejected Promise if those features do not seem to be available.

  - It requires the SharedArrayBuffer and Atomics classes, and the
    former is only available if the HTTP server emits the so-called
    COOP and COEP response headers. These features are required for
    proxying OPFS's synchronous API via the synchronous interface
    required by the sqlite3_vfs API.

  - This function may only be called a single time. When called, this
    function removes itself from the sqlite3 object.

  All arguments to this function are for internal/development purposes
  only. They do not constitute a public API and may change at any
  time.

  The argument may optionally be a plain object with the following
  configuration options:

  - proxyUri: as described above

  - verbose (=2): an integer 0-3. 0 disables all logging, 1 enables
    logging of errors. 2 enables logging of warnings and errors. 3
    additionally enables debugging info.

  - sanityChecks (=false): if true, some basic sanity tests are
    run on the OPFS VFS API after it's initialized, before the
    returned Promise resolves.

  On success, the Promise resolves to the top-most sqlite3 namespace
  object and that object gets a new object installed in its
  `opfs` property, containing several OPFS-specific utilities.
*/
const installOpfsVfs = function callee(options){
  if(!self.SharedArrayBuffer ||
     !self.Atomics ||
     !self.FileSystemHandle ||
     !self.FileSystemDirectoryHandle ||
     !self.FileSystemFileHandle ||
     !self.FileSystemFileHandle.prototype.createSyncAccessHandle ||
     !navigator.storage.getDirectory){
    return Promise.reject(
      new Error("This environment does not have OPFS support.")
    );
  }
  if(!options || 'object'!==typeof options){
    options = Object.create(null);
  }
  const urlParams = new URL(self.location.href).searchParams;
  if(undefined===options.verbose){
    options.verbose = urlParams.has('opfs-verbose') ? 3 : 2;
  }
  if(undefined===options.sanityChecks){
    options.sanityChecks = urlParams.has('opfs-sanity-check');
  }
  if(undefined===options.proxyUri){
    options.proxyUri = callee.defaultProxyUri;
  }

  if('function' === typeof options.proxyUri){
    options.proxyUri = options.proxyUri();
  }
  const thePromise = new Promise(function(promiseResolve, promiseReject_){
    const loggers = {
      0:console.error.bind(console),
      1:console.warn.bind(console),
      2:console.log.bind(console)
    };
    const logImpl = (level,...args)=>{
      if(options.verbose>level) loggers[level]("OPFS syncer:",...args);
    };
    const log =    (...args)=>logImpl(2, ...args);
    const warn =   (...args)=>logImpl(1, ...args);
    const error =  (...args)=>logImpl(0, ...args);
    const toss = function(...args){throw new Error(args.join(' '))};
    const capi = sqlite3.capi;
    const wasm = capi.wasm;
    const sqlite3_vfs = capi.sqlite3_vfs;
    const sqlite3_file = capi.sqlite3_file;
    const sqlite3_io_methods = capi.sqlite3_io_methods;
    /**
       Generic utilities for working with OPFS. This will get filled out
       by the Promise setup and, on success, installed as sqlite3.opfs.
    */
    const opfsUtil = Object.create(null);
    /**
       Not part of the public API. Solely for internal/development
       use.
    */
    opfsUtil.metrics = {
      dump: function(){
        let k, n = 0, t = 0, w = 0;
        for(k in state.opIds){
          const m = metrics[k];
          n += m.count;
          t += m.time;
          w += m.wait;
          m.avgTime = (m.count && m.time) ? (m.time / m.count) : 0;
          m.avgWait = (m.count && m.wait) ? (m.wait / m.count) : 0;
        }
        console.log(self.location.href,
                    "metrics for",self.location.href,":",metrics,
                    "\nTotal of",n,"op(s) for",t,
                    "ms (incl. "+w+" ms of waiting on the async side)");
        console.log("Serialization metrics:",metrics.s11n);
        W.postMessage({type:'opfs-async-metrics'});
      },
      reset: function(){
        let k;
        const r = (m)=>(m.count = m.time = m.wait = 0);
        for(k in state.opIds){
          r(metrics[k] = Object.create(null));
        }
        let s = metrics.s11n = Object.create(null);
        s = s.serialize = Object.create(null);
        s.count = s.time = 0;
        s = metrics.s11n.deserialize = Object.create(null);
        s.count = s.time = 0;
      }
    }/*metrics*/;      
    const promiseReject = function(err){
      opfsVfs.dispose();
      return promiseReject_(err);
    };
    const W = new Worker(options.proxyUri);
    W._originalOnError = W.onerror /* will be restored later */;
    W.onerror = function(err){
      // The error object doesn't contain any useful info when the
      // failure is, e.g., that the remote script is 404.
      error("Error initializing OPFS asyncer:",err);
      promiseReject(new Error("Loading OPFS async Worker failed for unknown reasons."));
    };
    const pDVfs = capi.sqlite3_vfs_find(null)/*pointer to default VFS*/;
    const dVfs = pDVfs
          ? new sqlite3_vfs(pDVfs)
          : null /* dVfs will be null when sqlite3 is built with
                    SQLITE_OS_OTHER. Though we cannot currently handle
                    that case, the hope is to eventually be able to. */;
    const opfsVfs = new sqlite3_vfs();
    const opfsIoMethods = new sqlite3_io_methods();
    opfsVfs.$iVersion = 2/*yes, two*/;
    opfsVfs.$szOsFile = capi.sqlite3_file.structInfo.sizeof;
    opfsVfs.$mxPathname = 1024/*sure, why not?*/;
    opfsVfs.$zName = wasm.allocCString("opfs");
    // All C-side memory of opfsVfs is zeroed out, but just to be explicit:
    opfsVfs.$xDlOpen = opfsVfs.$xDlError = opfsVfs.$xDlSym = opfsVfs.$xDlClose = null;
    opfsVfs.ondispose = [
      '$zName', opfsVfs.$zName,
      'cleanup default VFS wrapper', ()=>(dVfs ? dVfs.dispose() : null),
      'cleanup opfsIoMethods', ()=>opfsIoMethods.dispose()
    ];
    /**
       Pedantic sidebar about opfsVfs.ondispose: the entries in that array
       are items to clean up when opfsVfs.dispose() is called, but in this
       environment it will never be called. The VFS instance simply
       hangs around until the WASM module instance is cleaned up. We
       "could" _hypothetically_ clean it up by "importing" an
       sqlite3_os_end() impl into the wasm build, but the shutdown order
       of the wasm engine and the JS one are undefined so there is no
       guaranty that the opfsVfs instance would be available in one
       environment or the other when sqlite3_os_end() is called (_if_ it
       gets called at all in a wasm build, which is undefined).
    */
    /**
       State which we send to the async-api Worker or share with it.
       This object must initially contain only cloneable or sharable
       objects. After the worker's "inited" message arrives, other types
       of data may be added to it.

       For purposes of Atomics.wait() and Atomics.notify(), we use a
       SharedArrayBuffer with one slot reserved for each of the API
       proxy's methods. The sync side of the API uses Atomics.wait()
       on the corresponding slot and the async side uses
       Atomics.notify() on that slot.

       The approach of using a single SAB to serialize comms for all
       instances might(?) lead to deadlock situations in multi-db
       cases. We should probably have one SAB here with a single slot
       for locking a per-file initialization step and then allocate a
       separate SAB like the above one for each file. That will
       require a bit of acrobatics but should be feasible. The most
       problematic part is that xOpen() would have to use
       postMessage() to communicate its SharedArrayBuffer, and mixing
       that approach with Atomics.wait/notify() gets a bit messy.
    */
    const state = Object.create(null);
    state.verbose = options.verbose;
    state.littleEndian = (()=>{
      const buffer = new ArrayBuffer(2);
      new DataView(buffer).setInt16(0, 256, true /* ==>littleEndian */);
      // Int16Array uses the platform's endianness.
      return new Int16Array(buffer)[0] === 256;
    })();
    /**
       Whether the async counterpart should log exceptions to
       the serialization channel. That produces a great deal of
       noise for seemingly innocuous things like xAccess() checks
       for missing files, so this option may have one of 3 values:

       0 = no exception logging

       1 = only log exceptions for "significant" ops like xOpen(),
       xRead(), and xWrite().

       2 = log all exceptions.
    */
    state.asyncS11nExceptions = 1;
    /* Size of file I/O buffer block. 64k = max sqlite3 page size, and
       xRead/xWrite() will never deal in blocks larger than that. */
    state.fileBufferSize = 1024 * 64;
    state.sabS11nOffset = state.fileBufferSize;
    /**
       The size of the block in our SAB for serializing arguments and
       result values. Needs to be large enough to hold serialized
       values of any of the proxied APIs. Filenames are the largest
       part but are limited to opfsVfs.$mxPathname bytes.
    */
    state.sabS11nSize = opfsVfs.$mxPathname * 2;
    /**
       The SAB used for all data I/O between the synchronous and
       async halves (file i/o and arg/result s11n).
    */
    state.sabIO = new SharedArrayBuffer(
      state.fileBufferSize/* file i/o block */
      + state.sabS11nSize/* argument/result serialization block */
    );
    state.opIds = Object.create(null);
    const metrics = Object.create(null);
    {
      /* Indexes for use in our SharedArrayBuffer... */
      let i = 0;
      /* SAB slot used to communicate which operation is desired
         between both workers. This worker writes to it and the other
         listens for changes. */
      state.opIds.whichOp = i++;
      /* Slot for storing return values. This worker listens to that
         slot and the other worker writes to it. */
      state.opIds.rc = i++;
      /* Each function gets an ID which this worker writes to
         the whichOp slot. The async-api worker uses Atomic.wait()
         on the whichOp slot to figure out which operation to run
         next. */
      state.opIds.xAccess = i++;
      state.opIds.xClose = i++;
      state.opIds.xDelete = i++;
      state.opIds.xDeleteNoWait = i++;
      state.opIds.xFileControl = i++;
      state.opIds.xFileSize = i++;
      state.opIds.xLock = i++;
      state.opIds.xOpen = i++;
      state.opIds.xRead = i++;
      state.opIds.xSleep = i++;
      state.opIds.xSync = i++;
      state.opIds.xTruncate = i++;
      state.opIds.xUnlock = i++;
      state.opIds.xWrite = i++;
      state.opIds.mkdir = i++;
      state.opIds['opfs-async-metrics'] = i++;
      state.opIds['opfs-async-shutdown'] = i++;
      /* The retry slot is used by the async part for wait-and-retry
         semantics. Though we could hypothetically use the xSleep slot
         for that, doing so might lead to undesired side effects. */
      state.opIds.retry = i++;
      state.sabOP = new SharedArrayBuffer(
        i * 4/* ==sizeof int32, noting that Atomics.wait() and friends
                can only function on Int32Array views of an SAB. */);
      opfsUtil.metrics.reset();
    }
    /**
       SQLITE_xxx constants to export to the async worker
       counterpart...
    */
    state.sq3Codes = Object.create(null);
    [
      'SQLITE_ERROR', 'SQLITE_IOERR',
      'SQLITE_NOTFOUND', 'SQLITE_MISUSE',
      'SQLITE_IOERR_READ', 'SQLITE_IOERR_SHORT_READ',
      'SQLITE_IOERR_WRITE', 'SQLITE_IOERR_FSYNC',
      'SQLITE_IOERR_TRUNCATE', 'SQLITE_IOERR_DELETE',
      'SQLITE_IOERR_ACCESS', 'SQLITE_IOERR_CLOSE',
      'SQLITE_IOERR_DELETE',
      'SQLITE_LOCK_NONE',
      'SQLITE_LOCK_SHARED',
      'SQLITE_LOCK_RESERVED',
      'SQLITE_LOCK_PENDING',
      'SQLITE_LOCK_EXCLUSIVE',
      'SQLITE_OPEN_CREATE', 'SQLITE_OPEN_DELETEONCLOSE',
      'SQLITE_OPEN_READONLY'
    ].forEach((k)=>{
      if(undefined === (state.sq3Codes[k] = capi[k])){
        toss("Maintenance required: not found:",k);
      }
    });

    /**
       Runs the given operation (by name) in the async worker
       counterpart, waits for its response, and returns the result
       which the async worker writes to SAB[state.opIds.rc]. The
       2nd and subsequent arguments must be the aruguments for the
       async op.
    */
    const opRun = (op,...args)=>{
      const opNdx = state.opIds[op] || toss("Invalid op ID:",op);
      state.s11n.serialize(...args);
      Atomics.store(state.sabOPView, state.opIds.rc, -1);
      Atomics.store(state.sabOPView, state.opIds.whichOp, opNdx);
      Atomics.notify(state.sabOPView, state.opIds.whichOp)
      /* async thread will take over here */;
      const t = performance.now();
      Atomics.wait(state.sabOPView, state.opIds.rc, -1)
      /* When this wait() call returns, the async half will have
         completed the operation and reported its results. */;
      const rc = Atomics.load(state.sabOPView, state.opIds.rc);
      metrics[op].wait += performance.now() - t;
      if(rc && state.asyncS11nExceptions){
        const err = state.s11n.deserialize();
        if(err) error(op+"() async error:",...err);
      }
      return rc;
    };

    const initS11n = ()=>{
      /**
         !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
         ACHTUNG: this code is 100% duplicated in the other half of
         this proxy! The documentation is maintained in the
         "synchronous half".
         !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

         This proxy de/serializes cross-thread function arguments and
         output-pointer values via the state.sabIO SharedArrayBuffer,
         using the region defined by (state.sabS11nOffset,
         state.sabS11nOffset]. Only one dataset is recorded at a time.

         This is not a general-purpose format. It only supports the
         range of operations, and data sizes, needed by the
         sqlite3_vfs and sqlite3_io_methods operations. Serialized
         data are transient and this serialization algorithm may
         change at any time.

         The data format can be succinctly summarized as:

         Nt...Td...D

         Where:

         - N = number of entries (1 byte)

         - t = type ID of first argument (1 byte)

         - ...T = type IDs of the 2nd and subsequent arguments (1 byte
         each).

         - d = raw bytes of first argument (per-type size).

         - ...D = raw bytes of the 2nd and subsequent arguments (per-type
         size).

         All types except strings have fixed sizes. Strings are stored
         using their TextEncoder/TextDecoder representations. It would
         arguably make more sense to store them as Int16Arrays of
         their JS character values, but how best/fastest to get that
         in and out of string form is an open point. Initial
         experimentation with that approach did not gain us any speed.

         Historical note: this impl was initially about 1% this size by
         using using JSON.stringify/parse(), but using fit-to-purpose
         serialization saves considerable runtime.
      */
      if(state.s11n) return state.s11n;
      const textDecoder = new TextDecoder(),
            textEncoder = new TextEncoder('utf-8'),
            viewU8 = new Uint8Array(state.sabIO, state.sabS11nOffset, state.sabS11nSize),
            viewDV = new DataView(state.sabIO, state.sabS11nOffset, state.sabS11nSize);
      state.s11n = Object.create(null);
      /* Only arguments and return values of these types may be
         serialized. This covers the whole range of types needed by the
         sqlite3_vfs API. */
      const TypeIds = Object.create(null);
      TypeIds.number  = { id: 1, size: 8, getter: 'getFloat64', setter: 'setFloat64' };
      TypeIds.bigint  = { id: 2, size: 8, getter: 'getBigInt64', setter: 'setBigInt64' };
      TypeIds.boolean = { id: 3, size: 4, getter: 'getInt32', setter: 'setInt32' };
      TypeIds.string =  { id: 4 };

      const getTypeId = (v)=>(
        TypeIds[typeof v]
          || toss("Maintenance required: this value type cannot be serialized.",v)
      );
      const getTypeIdById = (tid)=>{
        switch(tid){
            case TypeIds.number.id: return TypeIds.number;
            case TypeIds.bigint.id: return TypeIds.bigint;
            case TypeIds.boolean.id: return TypeIds.boolean;
            case TypeIds.string.id: return TypeIds.string;
            default: toss("Invalid type ID:",tid);
        }
      };

      /**
         Returns an array of the deserialized state stored by the most
         recent serialize() operation (from from this thread or the
         counterpart thread), or null if the serialization buffer is empty.
      */
      state.s11n.deserialize = function(){
        ++metrics.s11n.deserialize.count;
        const t = performance.now();
        const argc = viewU8[0];
        const rc = argc ? [] : null;
        if(argc){
          const typeIds = [];
          let offset = 1, i, n, v;
          for(i = 0; i < argc; ++i, ++offset){
            typeIds.push(getTypeIdById(viewU8[offset]));
          }
          for(i = 0; i < argc; ++i){
            const t = typeIds[i];
            if(t.getter){
              v = viewDV[t.getter](offset, state.littleEndian);
              offset += t.size;
            }else{/*String*/
              n = viewDV.getInt32(offset, state.littleEndian);
              offset += 4;
              v = textDecoder.decode(viewU8.slice(offset, offset+n));
              offset += n;
            }
            rc.push(v);
          }
        }
        //log("deserialize:",argc, rc);
        metrics.s11n.deserialize.time += performance.now() - t;
        return rc;
      };

      /**
         Serializes all arguments to the shared buffer for consumption
         by the counterpart thread.

         This routine is only intended for serializing OPFS VFS
         arguments and (in at least one special case) result values,
         and the buffer is sized to be able to comfortably handle
         those.

         If passed no arguments then it zeroes out the serialization
         state.
      */
      state.s11n.serialize = function(...args){
        const t = performance.now();
        ++metrics.s11n.serialize.count;
        if(args.length){
          //log("serialize():",args);
          const typeIds = [];
          let i = 0, offset = 1;
          viewU8[0] = args.length & 0xff /* header = # of args */;
          for(; i < args.length; ++i, ++offset){
            /* Write the TypeIds.id value into the next args.length
               bytes. */
            typeIds.push(getTypeId(args[i]));
            viewU8[offset] = typeIds[i].id;
          }
          for(i = 0; i < args.length; ++i) {
            /* Deserialize the following bytes based on their
               corresponding TypeIds.id from the header. */
            const t = typeIds[i];
            if(t.setter){
              viewDV[t.setter](offset, args[i], state.littleEndian);
              offset += t.size;
            }else{/*String*/
              const s = textEncoder.encode(args[i]);
              viewDV.setInt32(offset, s.byteLength, state.littleEndian);
              offset += 4;
              viewU8.set(s, offset);
              offset += s.byteLength;
            }
          }
          //log("serialize() result:",viewU8.slice(0,offset));
        }else{
          viewU8[0] = 0;
        }
        metrics.s11n.serialize.time += performance.now() - t;
      };
      return state.s11n;
    }/*initS11n()*/;

    /**
       Generates a random ASCII string len characters long, intended for
       use as a temporary file name.
    */
    const randomFilename = function f(len=16){
      if(!f._chars){
        f._chars = "abcdefghijklmnopqrstuvwxyz"+
          "ABCDEFGHIJKLMNOPQRSTUVWXYZ"+
          "012346789";
        f._n = f._chars.length;
      }
      const a = [];
      let i = 0;
      for( ; i < len; ++i){
        const ndx = Math.random() * (f._n * 64) % f._n | 0;
        a[i] = f._chars[ndx];
      }
      return a.join('');
    };

    /**
       Map of sqlite3_file pointers to objects constructed by xOpen().
    */
    const __openFiles = Object.create(null);

    /**
       Installs a StructBinder-bound function pointer member of the
       given name and function in the given StructType target object.
       It creates a WASM proxy for the given function and arranges for
       that proxy to be cleaned up when tgt.dispose() is called.  Throws
       on the slightest hint of error (e.g. tgt is-not-a StructType,
       name does not map to a struct-bound member, etc.).

       Returns a proxy for this function which is bound to tgt and takes
       2 args (name,func). That function returns the same thing,
       permitting calls to be chained.

       If called with only 1 arg, it has no side effects but returns a
       func with the same signature as described above.
    */
    const installMethod = function callee(tgt, name, func){
      if(!(tgt instanceof sqlite3.StructBinder.StructType)){
        toss("Usage error: target object is-not-a StructType.");
      }
      if(1===arguments.length){
        return (n,f)=>callee(tgt,n,f);
      }
      if(!callee.argcProxy){
        callee.argcProxy = function(func,sig){
          return function(...args){
            if(func.length!==arguments.length){
              toss("Argument mismatch. Native signature is:",sig);
            }
            return func.apply(this, args);
          }
        };
        callee.removeFuncList = function(){
          if(this.ondispose.__removeFuncList){
            this.ondispose.__removeFuncList.forEach(
              (v,ndx)=>{
                if('number'===typeof v){
                  try{wasm.uninstallFunction(v)}
                  catch(e){/*ignore*/}
                }
                /* else it's a descriptive label for the next number in
                   the list. */
              }
            );
            delete this.ondispose.__removeFuncList;
          }
        };
      }/*static init*/
      const sigN = tgt.memberSignature(name);
      if(sigN.length<2){
        toss("Member",name," is not a function pointer. Signature =",sigN);
      }
      const memKey = tgt.memberKey(name);
      const fProxy = 0
      /** This middle-man proxy is only for use during development, to
          confirm that we always pass the proper number of
          arguments. We know that the C-level code will always use the
          correct argument count. */
            ? callee.argcProxy(func, sigN)
            : func;
      const pFunc = wasm.installFunction(fProxy, tgt.memberSignature(name, true));
      tgt[memKey] = pFunc;
      if(!tgt.ondispose) tgt.ondispose = [];
      if(!tgt.ondispose.__removeFuncList){
        tgt.ondispose.push('ondispose.__removeFuncList handler',
                           callee.removeFuncList);
        tgt.ondispose.__removeFuncList = [];
      }
      tgt.ondispose.__removeFuncList.push(memKey, pFunc);
      return (n,f)=>callee(tgt, n, f);
    }/*installMethod*/;

    const opTimer = Object.create(null);
    opTimer.op = undefined;
    opTimer.start = undefined;
    const mTimeStart = (op)=>{
      opTimer.start = performance.now();
      opTimer.op = op;
      ++metrics[op].count;
    };
    const mTimeEnd = ()=>(
      metrics[opTimer.op].time += performance.now() - opTimer.start
    );

    /**
       Impls for the sqlite3_io_methods methods. Maintenance reminder:
       members are in alphabetical order to simplify finding them.
    */
    const ioSyncWrappers = {
      xCheckReservedLock: function(pFile,pOut){
        /**
           As of late 2022, only a single lock can be held on an OPFS
           file. We have no way of checking whether any _other_ db
           connection has a lock except by trying to obtain and (on
           success) release a sync-handle for it, but doing so would
           involve an inherent race condition. For the time being,
           pending a better solution, we simply report whether the
           given pFile instance has a lock.
        */
        const f = __openFiles[pFile];
        wasm.setMemValue(pOut, f.lockMode ? 1 : 0, 'i32');
        return 0;
      },
      xClose: function(pFile){
        mTimeStart('xClose');
        let rc = 0;
        const f = __openFiles[pFile];
        if(f){
          delete __openFiles[pFile];
          rc = opRun('xClose', pFile);
          if(f.sq3File) f.sq3File.dispose();
        }
        mTimeEnd();
        return rc;
      },
      xDeviceCharacteristics: function(pFile){
        //debug("xDeviceCharacteristics(",pFile,")");
        return capi.SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN;
      },
      xFileControl: function(pFile, opId, pArg){
        mTimeStart('xFileControl');
        const rc = (capi.SQLITE_FCNTL_SYNC===opId)
              ? opRun('xSync', pFile, 0)
              : capi.SQLITE_NOTFOUND;
        mTimeEnd();
        return rc;
      },
      xFileSize: function(pFile,pSz64){
        mTimeStart('xFileSize');
        const rc = opRun('xFileSize', pFile);
        if(0==rc){
          const sz = state.s11n.deserialize()[0];
          wasm.setMemValue(pSz64, sz, 'i64');
        }
        mTimeEnd();
        return rc;
      },
      xLock: function(pFile,lockType){
        mTimeStart('xLock');
        const f = __openFiles[pFile];
        let rc = 0;
        if( capi.SQLITE_LOCK_NONE === f.lockType ) {
          rc = opRun('xLock', pFile, lockType);
          if( 0===rc ) f.lockType = lockType;
        }else{
          f.lockType = lockType;
        }
        mTimeEnd();
        return rc;
      },
      xRead: function(pFile,pDest,n,offset64){
        mTimeStart('xRead');
        const f = __openFiles[pFile];
        let rc;
        try {
          rc = opRun('xRead',pFile, n, Number(offset64));
          if(0===rc || capi.SQLITE_IOERR_SHORT_READ===rc){ 
            /**
               Results get written to the SharedArrayBuffer f.sabView.
               Because the heap is _not_ a SharedArrayBuffer, we have
               to copy the results. TypedArray.set() seems to be the
               fastest way to copy this. */
            wasm.heap8u().set(f.sabView.subarray(0, n), pDest);
          }
        }catch(e){
          error("xRead(",arguments,") failed:",e,f);
          rc = capi.SQLITE_IOERR_READ;
        }
        mTimeEnd();
        return rc;
      },
      xSync: function(pFile,flags){
        ++metrics.xSync.count;
        return 0; // impl'd in xFileControl()
      },
      xTruncate: function(pFile,sz64){
        mTimeStart('xTruncate');
        const rc = opRun('xTruncate', pFile, Number(sz64));
        mTimeEnd();
        return rc;
      },
      xUnlock: function(pFile,lockType){
        mTimeStart('xUnlock');
        const f = __openFiles[pFile];
        let rc = 0;
        if( capi.SQLITE_LOCK_NONE === lockType
          && f.lockType ){
          rc = opRun('xUnlock', pFile, lockType);
        }
        if( 0===rc ) f.lockType = lockType;
        mTimeEnd();
        return rc;
      },
      xWrite: function(pFile,pSrc,n,offset64){
        mTimeStart('xWrite');
        const f = __openFiles[pFile];
        let rc;
        try {
          f.sabView.set(wasm.heap8u().subarray(pSrc, pSrc+n));
          rc = opRun('xWrite', pFile, n, Number(offset64));
        }catch(e){
          error("xWrite(",arguments,") failed:",e,f);
          rc = capi.SQLITE_IOERR_WRITE;
        }
        mTimeEnd();
        return rc;
      }
    }/*ioSyncWrappers*/;

    /**
       Impls for the sqlite3_vfs methods. Maintenance reminder: members
       are in alphabetical order to simplify finding them.
    */
    const vfsSyncWrappers = {
      xAccess: function(pVfs,zName,flags,pOut){
        mTimeStart('xAccess');
        const rc = opRun('xAccess', wasm.cstringToJs(zName));
        wasm.setMemValue( pOut, (rc ? 0 : 1), 'i32' );
        mTimeEnd();
        return 0;
      },
      xCurrentTime: function(pVfs,pOut){
        /* If it turns out that we need to adjust for timezone, see:
           https://stackoverflow.com/a/11760121/1458521 */
        wasm.setMemValue(pOut, 2440587.5 + (new Date().getTime()/86400000),
                         'double');
        return 0;
      },
      xCurrentTimeInt64: function(pVfs,pOut){
        // TODO: confirm that this calculation is correct
        wasm.setMemValue(pOut, (2440587.5 * 86400000) + new Date().getTime(),
                         'i64');
        return 0;
      },
      xDelete: function(pVfs, zName, doSyncDir){
        mTimeStart('xDelete');
        opRun('xDelete', wasm.cstringToJs(zName), doSyncDir, false);
        /* We're ignoring errors because we cannot yet differentiate
           between harmless and non-harmless failures. */
        mTimeEnd();
        return 0;
      },
      xFullPathname: function(pVfs,zName,nOut,pOut){
        /* Until/unless we have some notion of "current dir"
           in OPFS, simply copy zName to pOut... */
        const i = wasm.cstrncpy(pOut, zName, nOut);
        return i<nOut ? 0 : capi.SQLITE_CANTOPEN
        /*CANTOPEN is required by the docs but SQLITE_RANGE would be a closer match*/;
      },
      xGetLastError: function(pVfs,nOut,pOut){
        /* TODO: store exception.message values from the async
           partner in a dedicated SharedArrayBuffer, noting that we'd have
           to encode them... TextEncoder can do that for us. */
        warn("OPFS xGetLastError() has nothing sensible to return.");
        return 0;
      },
      //xSleep is optionally defined below
      xOpen: function f(pVfs, zName, pFile, flags, pOutFlags){
        mTimeStart('xOpen');
        if(0===zName){
          zName = randomFilename();
        }else if('number'===typeof zName){
          zName = wasm.cstringToJs(zName);
        }
        const fh = Object.create(null);
        fh.fid = pFile;
        fh.filename = zName;
        fh.sab = new SharedArrayBuffer(state.fileBufferSize);
        fh.flags = flags;
        const rc = opRun('xOpen', pFile, zName, flags);
        if(!rc){
          /* Recall that sqlite3_vfs::xClose() will be called, even on
             error, unless pFile->pMethods is NULL. */
          if(fh.readOnly){
            wasm.setMemValue(pOutFlags, capi.SQLITE_OPEN_READONLY, 'i32');
          }
          __openFiles[pFile] = fh;
          fh.sabView = state.sabFileBufView;
          fh.sq3File = new sqlite3_file(pFile);
          fh.sq3File.$pMethods = opfsIoMethods.pointer;
          fh.lockType = capi.SQLITE_LOCK_NONE;
        }
        mTimeEnd();
        return rc;
      }/*xOpen()*/
    }/*vfsSyncWrappers*/;

    if(dVfs){
      opfsVfs.$xRandomness = dVfs.$xRandomness;
      opfsVfs.$xSleep = dVfs.$xSleep;
    }
    if(!opfsVfs.$xRandomness){
      /* If the default VFS has no xRandomness(), add a basic JS impl... */
      vfsSyncWrappers.xRandomness = function(pVfs, nOut, pOut){
        const heap = wasm.heap8u();
        let i = 0;
        for(; i < nOut; ++i) heap[pOut + i] = (Math.random()*255000) & 0xFF;
        return i;
      };
    }
    if(!opfsVfs.$xSleep){
      /* If we can inherit an xSleep() impl from the default VFS then
         assume it's sane and use it, otherwise install a JS-based
         one. */
      vfsSyncWrappers.xSleep = function(pVfs,ms){
        Atomics.wait(state.sabOPView, state.opIds.xSleep, 0, ms);
        return 0;
      };
    }

    /* Install the vfs/io_methods into their C-level shared instances... */
    for(let k of Object.keys(ioSyncWrappers)){
      installMethod(opfsIoMethods, k, ioSyncWrappers[k]);
    }
    for(let k of Object.keys(vfsSyncWrappers)){
      installMethod(opfsVfs, k, vfsSyncWrappers[k]);
    }

    /**
       Syncronously deletes the given OPFS filesystem entry, ignoring
       any errors. As this environment has no notion of "current
       directory", the given name must be an absolute path. If the 2nd
       argument is truthy, deletion is recursive (use with caution!).

       Returns true if the deletion succeeded and false if it fails,
       but cannot report the nature of the failure.
    */
    opfsUtil.deleteEntry = function(fsEntryName,recursive=false){
      mTimeStart('xDelete');
      const rc = opRun('xDelete', fsEntryName, 0, recursive);
      mTimeEnd();
      return 0===rc;
    };
    /**
       Synchronously creates the given directory name, recursively, in
       the OPFS filesystem. Returns true if it succeeds or the
       directory already exists, else false.
    */
    opfsUtil.mkdir = function(absDirName){
      mTimeStart('mkdir');
      const rc = opRun('mkdir', absDirName);
      mTimeEnd();
      return 0===rc;
    };
    /**
       Synchronously checks whether the given OPFS filesystem exists,
       returning true if it does, false if it doesn't.
    */
    opfsUtil.entryExists = function(fsEntryName){
      return 0===opRun('xAccess', fsEntryName);
    };

    /**
       Generates a random ASCII string, intended for use as a
       temporary file name. Its argument is the length of the string,
       defaulting to 16.
    */
    opfsUtil.randomFilename = randomFilename;

    /**
       Re-registers the OPFS VFS. This is intended only for odd use
       cases which have to call sqlite3_shutdown() as part of their
       initialization process, which will unregister the VFS
       registered by installOpfsVfs(). If passed a truthy value, the
       OPFS VFS is registered as the default VFS, else it is not made
       the default. Returns the result of the the
       sqlite3_vfs_register() call.

       Design note: the problem of having to re-register things after
       a shutdown/initialize pair is more general. How to best plug
       that in to the library is unclear. In particular, we cannot
       hook in to any C-side calls to sqlite3_initialize(), so we
       cannot add an after-initialize callback mechanism.
    */
    opfsUtil.registerVfs = (asDefault=false)=>{
      return capi.wasm.exports.sqlite3_vfs_register(
        opfsVfs.pointer, asDefault ? 1 : 0
      );
    };

    /**
       Only for test/development use.
    */
    opfsUtil.debug = {
      asyncShutdown: ()=>{
        warn("Shutting down OPFS async listener. OPFS will no longer work.");
        opRun('opfs-async-shutdown');
      },
      asyncRestart: ()=>{
        warn("Attempting to restart OPFS async listener. Might work, might not.");
        W.postMessage({type: 'opfs-async-restart'});
      }
    };

    //TODO to support fiddle db upload:
    //opfsUtil.createFile = function(absName, content=undefined){...}

    if(sqlite3.oo1){
      opfsUtil.OpfsDb = function(...args){
        const opt = sqlite3.oo1.dbCtorHelper.normalizeArgs(...args);
        opt.vfs = opfsVfs.$zName;
        sqlite3.oo1.dbCtorHelper.call(this, opt);
      };
      opfsUtil.OpfsDb.prototype = Object.create(sqlite3.oo1.DB.prototype);
      sqlite3.oo1.dbCtorHelper.setVfsPostOpenSql(
        opfsVfs.pointer,
        [
          /* Truncate journal mode is faster than delete or wal for
             this vfs, per speedtest1. */
          "pragma journal_mode=truncate;"
          /*
            This vfs benefits hugely from cache on moderate/large
            speedtest1 --size 50 and --size 100 workloads. We currently
            rely on setting a non-default cache size when building
            sqlite3.wasm. If that policy changes, the cache can
            be set here.
          */
          //"pragma cache_size=-8388608;"
        ].join('')
      );
    }

    /**
       Potential TODOs:

       - Expose one or both of the Worker objects via opfsUtil and
         publish an interface for proxying the higher-level OPFS
         features like getting a directory listing.
    */
    const sanityCheck = function(){
      const scope = wasm.scopedAllocPush();
      const sq3File = new sqlite3_file();
      try{
        const fid = sq3File.pointer;
        const openFlags = capi.SQLITE_OPEN_CREATE
              | capi.SQLITE_OPEN_READWRITE
        //| capi.SQLITE_OPEN_DELETEONCLOSE
              | capi.SQLITE_OPEN_MAIN_DB;
        const pOut = wasm.scopedAlloc(8);
        const dbFile = "/sanity/check/file"+randomFilename(8);
        const zDbFile = wasm.scopedAllocCString(dbFile);
        let rc;
        state.s11n.serialize("This is ä string.");
        rc = state.s11n.deserialize();
        log("deserialize() says:",rc);
        if("This is ä string."!==rc[0]) toss("String d13n error.");
        vfsSyncWrappers.xAccess(opfsVfs.pointer, zDbFile, 0, pOut);
        rc = wasm.getMemValue(pOut,'i32');
        log("xAccess(",dbFile,") exists ?=",rc);
        rc = vfsSyncWrappers.xOpen(opfsVfs.pointer, zDbFile,
                                   fid, openFlags, pOut);
        log("open rc =",rc,"state.sabOPView[xOpen] =",
            state.sabOPView[state.opIds.xOpen]);
        if(0!==rc){
          error("open failed with code",rc);
          return;
        }
        vfsSyncWrappers.xAccess(opfsVfs.pointer, zDbFile, 0, pOut);
        rc = wasm.getMemValue(pOut,'i32');
        if(!rc) toss("xAccess() failed to detect file.");
        rc = ioSyncWrappers.xSync(sq3File.pointer, 0);
        if(rc) toss('sync failed w/ rc',rc);
        rc = ioSyncWrappers.xTruncate(sq3File.pointer, 1024);
        if(rc) toss('truncate failed w/ rc',rc);
        wasm.setMemValue(pOut,0,'i64');
        rc = ioSyncWrappers.xFileSize(sq3File.pointer, pOut);
        if(rc) toss('xFileSize failed w/ rc',rc);
        log("xFileSize says:",wasm.getMemValue(pOut, 'i64'));
        rc = ioSyncWrappers.xWrite(sq3File.pointer, zDbFile, 10, 1);
        if(rc) toss("xWrite() failed!");
        const readBuf = wasm.scopedAlloc(16);
        rc = ioSyncWrappers.xRead(sq3File.pointer, readBuf, 6, 2);
        wasm.setMemValue(readBuf+6,0);
        let jRead = wasm.cstringToJs(readBuf);
        log("xRead() got:",jRead);
        if("sanity"!==jRead) toss("Unexpected xRead() value.");
        if(vfsSyncWrappers.xSleep){
          log("xSleep()ing before close()ing...");
          vfsSyncWrappers.xSleep(opfsVfs.pointer,2000);
          log("waking up from xSleep()");
        }
        rc = ioSyncWrappers.xClose(fid);
        log("xClose rc =",rc,"sabOPView =",state.sabOPView);
        log("Deleting file:",dbFile);
        vfsSyncWrappers.xDelete(opfsVfs.pointer, zDbFile, 0x1234);
        vfsSyncWrappers.xAccess(opfsVfs.pointer, zDbFile, 0, pOut);
        rc = wasm.getMemValue(pOut,'i32');
        if(rc) toss("Expecting 0 from xAccess(",dbFile,") after xDelete().");
        warn("End of OPFS sanity checks.");
      }finally{
        sq3File.dispose();
        wasm.scopedAllocPop(scope);
      }
    }/*sanityCheck()*/;

    W.onmessage = function({data}){
      //log("Worker.onmessage:",data);
      switch(data.type){
          case 'opfs-async-loaded':
            /*Arrives as soon as the asyc proxy finishes loading.
              Pass our config and shared state on to the async worker.*/
            W.postMessage({type: 'opfs-async-init',args: state});
            break;
          case 'opfs-async-inited':{
            /*Indicates that the async partner has received the 'init'
              and has finished initializing, so the real work can
              begin...*/
            try {
              const rc = capi.sqlite3_vfs_register(opfsVfs.pointer, 0);
              if(rc){
                toss("sqlite3_vfs_register(OPFS) failed with rc",rc);
              }
              if(opfsVfs.pointer !== capi.sqlite3_vfs_find("opfs")){
                toss("BUG: sqlite3_vfs_find() failed for just-installed OPFS VFS");
              }
              capi.sqlite3_vfs_register.addReference(opfsVfs, opfsIoMethods);
              state.sabOPView = new Int32Array(state.sabOP);
              state.sabFileBufView = new Uint8Array(state.sabIO, 0, state.fileBufferSize);
              state.sabS11nView = new Uint8Array(state.sabIO, state.sabS11nOffset, state.sabS11nSize);
              initS11n();
              if(options.sanityChecks){
                warn("Running sanity checks because of opfs-sanity-check URL arg...");
                sanityCheck();
              }
              navigator.storage.getDirectory().then((d)=>{
                W.onerror = W._originalOnError;
                delete W._originalOnError;
                sqlite3.opfs = opfsUtil;
                opfsUtil.rootDirectory = d;
                log("End of OPFS sqlite3_vfs setup.", opfsVfs);
                promiseResolve(sqlite3);
              });
            }catch(e){
              error(e);
              promiseReject(e);
            }
            break;
          }
          default:
            promiseReject(e);
            error("Unexpected message from the async worker:",data);
            break;
      }/*switch(data.type)*/
    }/*W.onmessage()*/;
  })/*thePromise*/;
  return thePromise;
}/*installOpfsVfs()*/;
installOpfsVfs.defaultProxyUri =
  "sqlite3-opfs-async-proxy.js";
self.sqlite3ApiBootstrap.initializersAsync.push(async (sqlite3)=>{
  if(sqlite3.scriptInfo && !sqlite3.scriptInfo.isWorker){
    return;
  }
  try{
    let proxyJs = installOpfsVfs.defaultProxyUri;
    if(sqlite3.scriptInfo.sqlite3Dir){
      installOpfsVfs.defaultProxyUri =
        sqlite3.scriptInfo.sqlite3Dir + proxyJs;
      //console.warn("installOpfsVfs.defaultProxyUri =",installOpfsVfs.defaultProxyUri);
    }
    return installOpfsVfs().catch((e)=>{
      console.warn("Ignoring inability to install OPFS sqlite3_vfs:",e.message);
    });
  }catch(e){
    console.error("installOpfsVfs() exception:",e);
    throw e;
  }
});
}/*sqlite3ApiBootstrap.initializers.push()*/);