/* -*- c++ -*- */
/*
 * Copyright 2013, 2018 Free Software Foundation, Inc.
 *
 * This file is part of GNU Radio
 *
 * SPDX-License-Identifier: GPL-3.0-or-later
 *
 */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include "ofdm_cyclic_prefixer_impl.h"
#include <gnuradio/io_signature.h>

namespace gr {
namespace digital {

// Do not break backwards compatibility and overload the make function.
ofdm_cyclic_prefixer::sptr ofdm_cyclic_prefixer::make(size_t input_size,
                                                      size_t output_size,
                                                      int rolloff_len,
                                                      const std::string& len_tag_key)
{
    int fft_len = input_size;
    std::vector<int> cp_lengths(
        1, static_cast<int>(output_size - input_size)); // Cast to silence compiler :(
    return gnuradio::make_block_sptr<ofdm_cyclic_prefixer_impl>(
        fft_len, cp_lengths, rolloff_len, len_tag_key);
}

ofdm_cyclic_prefixer::sptr ofdm_cyclic_prefixer::make(int fft_len,
                                                      const std::vector<int>& cp_lengths,
                                                      int rolloff_len,
                                                      const std::string& len_tag_key)
{
    return gnuradio::make_block_sptr<ofdm_cyclic_prefixer_impl>(
        fft_len, cp_lengths, rolloff_len, len_tag_key);
}

ofdm_cyclic_prefixer_impl::ofdm_cyclic_prefixer_impl(int fft_len,
                                                     const std::vector<int>& cp_lengths,
                                                     int rolloff_len,
                                                     const std::string& len_tag_key)
    : gr::tagged_stream_block("ofdm_cyclic_prefixer",
                              gr::io_signature::make(1, 1, fft_len * sizeof(gr_complex)),
                              gr::io_signature::make(1, 1, sizeof(gr_complex)),
                              len_tag_key),
      d_fft_len(fft_len),
      d_state(0),
      d_cp_max(0),
      d_cp_min(std::numeric_limits<int>::max()),
      d_rolloff_len(rolloff_len),
      d_cp_lengths(cp_lengths),
      d_up_flank((rolloff_len ? rolloff_len - 1 : 0), 0),
      d_down_flank((rolloff_len ? rolloff_len - 1 : 0), 0),
      d_delay_line(0, 0),
      d_len_tag_key(len_tag_key)
{
    // Sanity
    if (d_cp_lengths.empty()) {
        throw std::invalid_argument(this->alias() +
                                    std::string(": CP lengths vector can not be empty."));
    }
    for (size_t i = 0; i < d_cp_lengths.size(); i++) {
        if (d_cp_lengths[i] != 0) {
            break;
        }
        if (i == d_cp_lengths.size() - 1) {
            throw std::invalid_argument(
                this->alias() +
                std::string(": Please provide at least one CP which is != 0."));
        }
    }
    for (const int cp_length : d_cp_lengths) {
        d_cp_max = std::max(d_cp_max, cp_length);
        d_cp_min = std::min(d_cp_min, cp_length);
    }
    if (d_cp_min < 0) {
        throw std::invalid_argument(this->alias() +
                                    std::string(": The minimum CP allowed is 0."));
    }
    // Give the buffer allocator and scheduler a hint about the ratio between input and
    // output.
    set_relative_rate(d_cp_max + d_fft_len);
    // Flank of length 1 would just be rectangular.
    if (d_rolloff_len == 1) {
        d_rolloff_len = 0;
        GR_LOG_WARN(d_logger,
                    "Set rolloff to 0, because 1 would result in a boxcar function.");
    }
    if (d_rolloff_len) {
        d_delay_line.resize(d_rolloff_len - 1, 0);
        // More sanity
        if (d_rolloff_len > d_cp_min) {
            throw std::invalid_argument(
                this->alias() + std::string(": Rolloff length must be smaller than any "
                                            "of the cyclic prefix lengths."));
        }
        /* The actual flanks are one sample shorter than d_rolloff_len, because the
           first sample of the up- and down flank is always zero and one, respectively.*/
        for (int i = 1; i < d_rolloff_len; i++) {
            d_up_flank[i - 1] = 0.5 * (1 + cos(M_PI * i / rolloff_len - M_PI));
            d_down_flank[i - 1] =
                0.5 * (1 + cos(M_PI * (rolloff_len - i) / rolloff_len - M_PI));
        }
    }
    if (d_len_tag_key.empty()) {
        // noutput_items is set to be a multiple of the largest possible output size.
        // It is always OK to return less (in case of the shorter CP).
        set_output_multiple(d_fft_len + d_cp_max);
    } else {
        // Avoid automatic tag propagation and propagate them manually.
        set_tag_propagation_policy(TPP_DONT);
    }
}

ofdm_cyclic_prefixer_impl::~ofdm_cyclic_prefixer_impl() {}

int ofdm_cyclic_prefixer_impl::calculate_output_stream_length(
    const gr_vector_int& ninput_items)
{
    int noutput_items = ninput_items[0] * (d_cp_max + d_fft_len) +
                        (d_len_tag_key.empty() ? 0 : d_delay_line.size());
    return noutput_items;
}


// Operates in two ways:
// - When there's a length tag name specified, operates in packet mode.
//   Here, an entire OFDM frame is processed at once. The final OFDM symbol
//   is postfixed with the delay line of the pulse shape.
//   We manually propagate tags.
// - Otherwise, we're in freewheeling mode. Process as many OFDM symbols as
//   are space for in the output buffer. The delay line is never flushed.
//   Tags are propagated by the scheduler.
int ofdm_cyclic_prefixer_impl::work(int noutput_items,
                                    gr_vector_int& ninput_items,
                                    gr_vector_const_void_star& input_items,
                                    gr_vector_void_star& output_items)
{
    gr_complex* in = (gr_complex*)input_items[0];
    gr_complex* out = (gr_complex*)output_items[0];
    int symbols_to_read = 0;
    // 1) Figure out if we're in freewheeling or packet mode.
    if (!d_len_tag_key.empty()) {
        symbols_to_read = ninput_items[0];
    } else {
        symbols_to_read =
            std::min(noutput_items / (int)(d_fft_len + d_cp_max), ninput_items[0]);
    }
    noutput_items = 0;
    // 2) Do the cyclic prefixing and, optionally, the pulse shaping.
    for (int sym_idx = 0; sym_idx < symbols_to_read; sym_idx++) {
        memcpy(static_cast<void*>(out + d_cp_lengths[d_state]),
               static_cast<void*>(in),
               d_fft_len * sizeof(gr_complex));
        memcpy(static_cast<void*>(out),
               static_cast<void*>(in + d_fft_len - d_cp_lengths[d_state]),
               d_cp_lengths[d_state] * sizeof(gr_complex));
        if (d_rolloff_len) {
            for (int i = 0; i < d_rolloff_len - 1; i++) {
                out[i] = out[i] * d_up_flank[i] + d_delay_line[i];
                /* This is basically a cyclic suffix, but completely shifted into the next
                   symbol. The data rate does not change. */
                d_delay_line[i] = in[i] * d_down_flank[i];
            }
        }
        in += d_fft_len;
        out += d_fft_len + d_cp_lengths[d_state];
        // Raise the number of noutput_items depending on how long the current output was.
        noutput_items += d_fft_len + d_cp_lengths[d_state];
        // Propagate tags.
        unsigned last_state = d_state > 0 ? d_state - 1 : d_cp_lengths.size() - 1;
        std::vector<tag_t> tags;
        get_tags_in_range(
            tags, 0, nitems_read(0) + sym_idx, nitems_read(0) + sym_idx + 1);
        for (unsigned i = 0; i < tags.size(); i++) {
            tags[i].offset = ((tags[i].offset - nitems_read(0)) *
                              (d_fft_len + d_cp_lengths[last_state])) +
                             nitems_written(0);
            add_item_tag(0, tags[i].offset, tags[i].key, tags[i].value);
        }
        // Finally switch to next state.
        ++d_state;
        d_state %= d_cp_lengths.size();
    }
    /* 3) If we're in packet mode:
          - flush the delay line, if applicable */
    if (!d_len_tag_key.empty()) {
        if (d_rolloff_len) {
            std::memcpy(static_cast<void*>(out),
                        static_cast<void*>(d_delay_line.data()),
                        sizeof(gr_complex) * d_delay_line.size());
            d_delay_line.assign(d_delay_line.size(), 0);
            // Make last symbol a bit longer.
            noutput_items += d_delay_line.size();
        }
    } else {
        consume_each(symbols_to_read);
    }

    return noutput_items;
}

} /* namespace digital */
} /* namespace gr */